system

Abstract

Provide a system. [Solution] means for acquiring video data from a monitoring device; means for pre-processing the acquired video data; means for transmitting the preprocessed video data to an analysis device; An analysis device analyzes the video data and detects abnormal behavior; Measures to anonymize the analysis results and protect personally identifiable information; a means for notifying the user terminal and related systems of the anonymized analysis results; A means of storing the analysis results in a database and using them as training data for the system; A system including a means for retraining the model of the analysis device based on new data to improve analysis accuracy.

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Although surveillance equipment is now widely used, video data analysis still faces challenges. Specifically, the high cost of implementing AI analysis has prevented widespread adoption, and privacy concerns remain when it comes to sharing video data. The present invention aims to address these challenges by providing a system that efficiently analyzes surveillance equipment video data, enhancing crime prevention while protecting privacy. [Means for solving the problem]

[0005] The present invention provides a system that includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect abnormal behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying user terminals and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as training data for the system, and means for retraining the model of the analysis device based on new data to improve analysis accuracy, thereby enhancing crime prevention effects and realizing privacy protection.

[0006] "Monitoring devices" are devices such as cameras and sensors used to acquire video data.

[0007] "Video data" refers to video feeds and image data captured by surveillance equipment.

[0008] "Preprocessing" refers to processes that make video data easier to analyze, such as noise removal and frame rate normalization.

[0009] An "analysis device" is hardware or software for analyzing pre-processed video data.

[0010] "Abnormal behavior" refers to behavior that is not normal and requires vigilance, such as intrusion or violent behavior.

[0011] "Anonymization" is the process of removing or protecting personally identifiable information.

[0012] A "user terminal" is a device used by a user, such as a smartphone or a personal computer.

[0013] "Relevant systems" are other systems or organizations that require the results of the analysis.

[0014] "Notification" refers to sending analysis results and alert information to user terminals and related systems.

[0015] A "database" is a storage device for storing analysis results and related information.

[0016] "Learning data" refers to data of past analysis results that are used to retrain the model of the analysis device.

[0017] "Retraining" is the process of improving the accuracy of an analyzer's model based on new data. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11]FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0019] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0020] First, the terms used in the following description will be explained.

[0021] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0022] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0023] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0024] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0025] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0026] [First embodiment]

[0027] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0028] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0029] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0030] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0031] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0032] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0033] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0034] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0035] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0036] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0037] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0038] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0039] A specific embodiment of a surveillance device video analysis system will be described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to prevent crime.

[0040] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0041] The analysis device then receives the preprocessed video data and analyzes it using deep learning models to detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0042] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, removing any personally identifiable information by pixelating faces and masking voices.

[0043] The server then notifies the appropriate user device and related systems of the analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take action. Additionally, the server can securely share anonymized data with relevant authorities (police, security companies) as needed.

[0044] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0045] Specific examples

[0046] Example of suspicious person detection in a commercial facility

[0047] Monitoring equipment installed in commercial facilities transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly.

[0048] In this way, surveillance video analysis systems can efficiently detect abnormal behavior and quickly respond to crime, thereby preventing crime. Furthermore, privacy protection is guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0049] The processing flow will be explained below.

[0050] Step 1:

[0051] The server acquires video data in real time from monitoring devices connected to the network. The monitoring devices come in a variety of forms, such as fixed cameras and dashcams, and the server connects to these devices to collect data.

[0052] Step 2:

[0053] The server temporarily stores the acquired video data in storage. This storage process is performed to back up the data so that it is not lost.

[0054] Step 3:

[0055] The server then begins pre-processing the video data stored in the storage, which includes noise removal, frame rate normalization, and resolution adjustment.

[0056] Step 4:

[0057] The server transmits the pre-processed video data to the analysis device, which transmits the data quickly and securely and ensures that the analysis device receives the data.

[0058] Step 5:

[0059] The analyzer receives the preprocessed data and uses deep learning models to analyze people's movements and abnormal behavior. In this process, deep learning algorithms detect abnormal behavior (such as intrusions or violent behavior) in the video.

[0060] Step 6:

[0061] When the analysis device detects abnormal behavior, it records the timestamp and location information and sends the analysis results to the server.

[0062] Step 7:

[0063] The server receives the analysis results, anonymizes any information that could identify individuals to protect privacy, and applies mosaic processing and audio masking to the video data.

[0064] Step 8:

[0065] The server notifies the user and related systems of the anonymized analysis results, and the user receives the notification on their smartphone or PC and can take action.

[0066] Step 9:

[0067] Users can check the notification and take immediate action if necessary. For example, a commercial facility manager can work with security staff to investigate the situation and take appropriate action.

[0068] Step 10:

[0069] The server stores the analysis results in a database, which will be used as learning data in the future.

[0070] Step 11:

[0071] The server periodically uses the stored data to retrain the analysis device's model, improving analysis accuracy and enabling more effective detection of abnormal behavior.

[0072] Step 12:

[0073] The server applies the retrained model to the system and continues analyzing the monitoring device data with the new analytical accuracy.

[0074] Example 1

[0075] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0076] Conventional surveillance systems have a problem in that they have low accuracy in analyzing video data, making it difficult to accurately detect abnormal behavior. Furthermore, they lack the functionality to detect and notify abnormal behavior in real time while protecting individual privacy, which often results in delayed response. To solve these problems, an efficient and highly accurate video data analysis system is required.

[0077] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0078] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for performing noise reduction and frame rate normalization on the preprocessed video data, means for recording timestamps and location information of anomalous behavior detected by the analysis device, means for performing face mosaic processing and audio masking when acquiring the analysis results and anonymizing the data for privacy protection, means for notifying the user terminal of the analysis results using an API or push notification system, and means for acquiring video data from the monitoring device in real time, preprocessing the video data, and transferring the video data to the analysis device, thereby enabling highly accurate detection of anomalous behavior.

[0079] A "monitoring device" is a device that continuously captures images of a monitored area and generates video data, and specifically includes fixed cameras and drive recorders.

[0080] "Video data" refers to video information acquired by a monitoring device and is made up of successive frames of images.

[0081] "Preprocessing" refers to processing such as noise removal and frame rate normalization before analyzing video data.

[0082] The "analysis device" is a device that runs a deep learning model to detect abnormal behavior using preprocessed video data as input.

[0083] "Abnormal behavior" refers to any unusual behavior or condition that occurs within the monitored area, including, for example, intrusion, violent behavior, or abnormal vehicle stalls.

[0084] "Anonymization" is a process carried out to protect information that could identify an individual from the analysis results, and specifically includes blurring faces and masking audio.

[0085] A "user terminal" is an electronic device capable of receiving and displaying analysis results, such as a smartphone or a personal computer.

[0086] A "database" is a system for efficiently storing and managing analysis results and learning data.

[0087] "Retraining" is the process of updating a deep learning model based on new data to improve analysis accuracy.

[0088] "Noise reduction" is the process of removing unnecessary background information and noise from video data.

[0089] "Frame rate normalization" is a process of standardizing the frame rate of video data to a fixed value.

[0090] A "timestamp" is information indicating a specific time, and is recorded in video data or analysis results.

[0091] "Location information" is information indicating the location where abnormal behavior was detected, and includes geographic coordinates, addresses, etc.

[0092] "API" stands for Application Program Interface, an interface that allows software functions and data to be used by other programs.

[0093] A "push notification system" is a system that sends notifications from a server to a user terminal in real time.

[0094] "Face mosaic processing" is a process of processing video data so that an individual's face cannot be identified, and includes polygon and blur effects.

[0095] "Audio masking" is a process of processing audio data to prevent individuals from being identified.

[0096] This invention describes a specific embodiment of a surveillance device video analysis system. This system is mainly composed of a server, a terminal, and a user, and aims to efficiently analyze video data acquired from surveillance devices and prevent crime.

[0097] First, the server acquires video data in real time from monitoring devices connected to the network (for example, fixed cameras or drive recorders). The server temporarily stores this video data in storage. To acquire the video data, for example, the video stream URL of an IP camera can be used.

[0098] The server then performs preprocessing on the acquired video data. This preprocessing includes noise reduction and frame rate normalization. For example, noise reduction using a Gaussian filter and standardizing the video frame rate to 30 fps are performed. This preprocessing allows for the efficient execution of subsequent analysis processes.

[0099] The server then sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the data and detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for abnormal behavior.

[0100] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy by removing any personally identifiable information, such as by blurring faces (e.g., using AWS® Rekognition) and masking voices.

[0101] The server then notifies the user's device (e.g., smartphone or PC) and related systems of the analysis results. Notifications are sent using APIs or push notification systems. Users can receive notifications on their devices and take action. If necessary, the server also securely shares anonymized data with related organizations (e.g., police or security companies).

[0102] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. For example, Tensorflow (registered trademark) or PyTorch can be used for this retraining.

[0103] Specific examples

[0104] Example of suspicious person detection in a commercial facility

[0105] Surveillance devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly. In this way, the surveillance device video analysis system efficiently detects abnormal behavior and acts quickly to deter crime. Privacy protection is also guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0106] Example prompts for generative AI models

[0107] text

[0108] Surveillance cameras installed in a commercial facility send video footage to a server in real time. The server preprocesses this video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the smartphone of the commercial facility manager. The manager receives the alert and responds promptly. Please explain this series of processes in easy-to-understand text.

[0109] This system makes it possible to efficiently analyze video data acquired from surveillance cameras, quickly detect and notify users of abnormal behavior, and provides a sense of security to users as it also takes privacy protection into consideration.

[0110] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0111] Step 1:

[0112] Acquiring and storing video data

[0113] The server acquires video data in real time from surveillance devices (e.g., fixed cameras or dashcams). This is done using the IP address or stream URL of the surveillance camera. The server then temporarily stores the acquired video data in storage. Specifically, it establishes a streaming connection for the video data and stores the acquired video frames sequentially.

[0114] Input: Surveillance video stream

[0115] Output: Temporarily saved video data

[0116] Step 2:

[0117] Video data preprocessing

[0118] The server performs preprocessing on the stored video data. This preprocessing includes noise reduction and frame rate normalization. A Gaussian filter is used for noise reduction, and frame rate normalization involves thinning and interpolating frames along the time axis. This improves data quality and increases the accuracy of analysis.

[0119] Input: Temporarily stored video data

[0120] Output: Pre-processed video data

[0121] Step 3:

[0122] Analysis of preprocessed video data

[0123] The server sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the video data and detect human movements and abnormal behavior. This analysis model operates in real time using pre-trained parameters.

[0124] Input: Preprocessed video data

[0125] Output: Analysis results (detection results of abnormal behavior, timestamp, location information)

[0126] Step 4:

[0127] Anonymization of analysis results

[0128] The server receives the analysis results and performs face blurring and audio masking to protect privacy. Face blurring is performed using image recognition software (e.g., AWS Rekognition), and audio masking is performed using technology that removes specific frequency bands.

[0129] Input: Analysis results

[0130] Output: Anonymized analysis results (face mosaic and voice masked data)

[0131] Step 5:

[0132] Notification to user devices and related systems

[0133] The server notifies the user's device and related systems of the anonymized analysis results using an API or push notification system. Users can receive the notifications on their smartphones or computers and respond promptly.

[0134] Input: Anonymized analysis results

[0135] Output: Notification to user device (push notification to smartphone or PC)

[0136] Step 6:

[0137] Saving analysis results and retraining models

[0138] The server stores the analysis results in a database and uses them as training data for the system. The analysis device model is periodically retrained based on new data to improve analysis accuracy. This retraining process uses TensorFlow and PyTorch to update the model parameters.

[0139] Input: Analysis results

[0140] Output: Updated analytical model, saved training data

[0141] These steps enable the system to achieve highly accurate detection of abnormal behavior, protect privacy, and support rapid response.

[0142] (Application example 1)

[0143] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0144] Conventional monitoring devices and systems often have delays in detecting abnormal behavior and subsequent responses, making it difficult to respond quickly. Another issue is that data privacy protection is insufficient, increasing the risk of personal information leaks. Furthermore, improving system training and analysis accuracy takes a lot of time and effort. There is a need for a system that can solve these issues and detect and respond to abnormal behavior more efficiently and quickly.

[0145] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0146] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect abnormal behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for notifying a smartphone application of abnormal behavior alerts in real time, and means for the user to view past abnormal behavior detection history. This enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases user peace of mind.

[0147] A "surveillance device" is a device for acquiring video data, such as a fixed camera or a drive recorder.

[0148] "Video data" refers to visual information acquired from a monitoring device, and is data used to detect abnormal behavior.

[0149] "Preprocessing" refers to the process of removing noise from the acquired video data, normalizing the frame rate, and otherwise processing the data to make it easier to analyze.

[0150] The "analysis device" is a device for analyzing pre-processed video data and detecting abnormal behavior.

[0151] "Abnormal behavior" refers to behavior that is different from the norm, such as the intrusion of suspicious individuals, violent behavior, or abnormal stopping of vehicles.

[0152] "Anonymization" is a process that removes information that can identify individuals from the analysis results in order to protect privacy.

[0153] A "user terminal" is a device, such as a smartphone or PC, that a user uses to check analysis results and receive notifications.

[0154] "Related systems" are external systems, such as police and security companies, that are necessary for sharing analysis results.

[0155] A "database" is a data storage device that stores analysis results and uses them as learning data for the system.

[0156] "Relearning" is the process of retraining the analysis device's model based on new data to improve analysis accuracy.

[0157] A "smartphone application" is software that runs on a smartphone and alerts users of abnormal behavior in real time.

[0158] An "abnormal behavior alert" is a warning or information that quickly notifies the user of detected abnormal behavior.

[0159] "History" is a record of abnormal behavior detected in the past, and is data including information that can be confirmed by the user.

[0160] A specific embodiment of a surveillance device video analysis system is described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from a surveillance device, detects abnormal behavior, and promptly notifies the user.

[0161] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored on storage (e.g., SSD or HDD). The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. For this purpose, libraries such as OpenCV are used. The processed video data is then sent to an analysis device.

[0162] The analysis device then receives the preprocessed video data and analyzes it using deep learning models (e.g., Keras or TensorFlow). This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0163] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. This includes blurring faces and masking audio to remove any personally identifiable information. The server then notifies the appropriate user device (e.g., smartphone, PC) and related systems (e.g., police, security companies) of the analysis results. Users receive alerts of abnormal behavior in real time through a smartphone application, allowing them to respond quickly. Users can also view past abnormal behavior detection history and consider countermeasures.

[0164] Furthermore, the server stores the analysis results in a database, which is used as training data for future systems. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. Retraining is performed using deep learning libraries such as Keras and TensorFlow.

[0165] As a specific example, consider a case where surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond quickly. In this way, the surveillance equipment video analysis system efficiently detects abnormal behavior and aims to deter crime by responding quickly.

[0166] Analysis using generative AI models uses prompts like the following:

[0167] "Analyze this video data to detect any abnormal behavior (e.g., unauthorized entry, violent behavior, or abnormal stopping). Include timestamps, detailed information, and location information in the detection results. Anonymize the data to protect privacy."

[0168] The above is a specific embodiment. This system enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases the sense of security of users.

[0169] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0170] Step 1:

[0171] The server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. The input is video data from the monitoring devices, and the output is data that is temporarily stored in storage. Specifically, the server receives video data sent from the monitoring devices and temporarily stores it.

[0172] Step 2:

[0173] The server performs preprocessing on the acquired video data, such as noise removal and frame rate normalization. The input here is the video data stored in storage, and the output is preprocessed, clean video data. Specifically, the server uses the OpenCV library to remove noise from the video and adjust the frame rate to a constant value.

[0174] Step 3:

[0175] The server sends the preprocessed video data to the analysis device. The input is the preprocessed video data, and the output is the data sent to the analysis device. Specifically, the preprocessed data is transferred to the analysis device via a network.

[0176] Step 4:

[0177] The analysis device receives the preprocessed video data and analyzes it using a deep learning model. This analysis detects abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The input is the preprocessed video data sent to the analysis device, and the output is the abnormal behavior detection results (timestamp, detailed information, and location information). Specifically, the data is analyzed using a generative AI model using Keras and TensorFlow.

[0178] Step 5:

[0179] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior detection results from the analysis device, and the output is anonymized data. Specific operations include face mosaic processing and voice masking to remove personally identifiable information.

[0180] Step 6:

[0181] The server notifies the user device (smartphone, PC) and related systems (police, security companies) of the anonymized analysis results. The input is the anonymized analysis results, and the output is the notified data. Specifically, the server sends the data to the user device or external system via an HTTP request, etc.

[0182] Step 7:

[0183] Users can receive alerts of abnormal behavior in real time through a smartphone application. The input is notification data from the server, and the output is alert information displayed on the smartphone application. Specifically, the application notification function receives the data and displays it on the screen.

[0184] Step 8:

[0185] Users can view the history of past abnormal behavior detections using a smartphone application. The input is the analysis results stored in the database, and the output is the history information displayed on the application. Specifically, the application queries the database, retrieves past detection results, and displays them.

[0186] Step 9:

[0187] The server stores the analysis results in a database and uses them as learning data for the system. The input is the analysis results, and the output is the data stored in the database. Specifically, the analysis results are formalized and stored in the database.

[0188] Step 10:

[0189] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is new data and past analysis results, and the output is the retrained model. Specifically, the model is updated using the retraining functions of Keras or TensorFlow.

[0190] This series of steps enables efficient and rapid detection and response to abnormal behavior, ensuring the safety of users.

[0191] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0192] Below, we will explain a specific embodiment that combines a surveillance video analysis system with an emotion engine. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to deter crime and recognize user emotions.

[0193] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0194] The analysis device then receives the preprocessed video data and uses deep learning models to analyze human movements and abnormal behaviors (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behaviors.

[0195] Furthermore, the analysis device is equipped with an emotion engine that analyzes the user's facial expressions and tone of voice from the video data to identify their emotions. This emotion analysis can determine whether the user is expressing emotions such as fear, surprise, or anger.

[0196] Once the analysis is complete, the server retrieves the results and anonymizes them to protect privacy. It uses techniques like face blurring and voice masking to remove any personally identifiable information. Sentiment analysis results are also anonymized.

[0197] The server then notifies the appropriate user device and related systems of the analysis and sentiment analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take appropriate action. Additionally, the server can securely share anonymized data with relevant agencies (police, security companies) as needed.

[0198] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0199] Specific examples

[0200] Example of suspicious person detection in a commercial facility

[0201] Monitoring devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, the emotion engine analyzes the video data to determine the suspicious person's fear or tension, and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0202] In this way, a system that combines a surveillance video analysis system and an emotion engine can efficiently detect abnormal behavior and quickly respond to it, thereby preventing crime. Emotion analysis also enables early detection of potential danger, further enhancing safety. Privacy protection is also guaranteed, improving users' sense of security. This system can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0203] The processing flow will be explained below.

[0204] Step 1:

[0205] The server receives video data in real time from network-connected surveillance devices, such as fixed cameras and dashcams, and connects to these devices to receive their signals.

[0206] Step 2:

[0207] The server temporarily stores the acquired video data in storage so that the data can be referenced later for processing.

[0208] Step 3:

[0209] The server performs preprocessing on the video data stored in the storage, including noise removal, frame rate normalization, and resolution adjustment.

[0210] Step 4:

[0211] The server transmits the pre-processed video data to the analysis equipment, and this transmission is fast and secure, ensuring that the data reaches the analysis equipment reliably.

[0212] Step 5:

[0213] The analysis device receives the preprocessed video data and uses deep learning models to analyze people and their behavior in the video, detecting abnormal behavior (such as intrusions, violent behavior, or abnormal vehicle stalls).

[0214] Step 6:

[0215] If the analysis device detects abnormal behavior, it records the timestamp and location information and sends this to the server as the analysis result.

[0216] Step 7:

[0217] The emotion engine included in the analysis device analyzes the user's facial expressions and tone of voice from the video data to generate emotion data, which can include fear, surprise, anger, etc.

[0218] Step 8:

[0219] The analysis device sends the emotion data to the server, where it is treated in the same way as the analysis results of abnormal behavior.

[0220] Step 9:

[0221] The server receives the analysis results and emotion data, and anonymizes them to protect privacy. It also applies mosaic processing to the video and masks the audio data.

[0222] Step 10:

[0223] The server notifies the user device and related systems of the anonymized analysis results and emotion data, and if notification is required, sends an alert to specific users or systems.

[0224] Step 11:

[0225] Users receive notifications on their smartphones or PCs and can take action as necessary. For example, a commercial facility manager can immediately contact security staff to check the situation and take appropriate measures.

[0226] Step 12:

[0227] The server stores the analysis results and emotion data in a database, which will be used as learning data for the system in the future.

[0228] Step 13:

[0229] The server periodically uses the saved data to retrain the analysis device's model, improving the model's analysis accuracy and enabling it to detect abnormal behavior and user emotions with greater accuracy.

[0230] Step 14:

[0231] The server applies the retrained model to the system and continues analyzing the video data from the surveillance device with improved analysis accuracy.

[0232] Example 2

[0233] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0234] Conventional surveillance systems were specialized in detecting abnormal behavior, but lacked the ability to analyze people's emotions, making it difficult to detect potential dangers early on. Furthermore, from the perspective of protecting privacy, it was necessary to adequately protect personally identifiable information.

[0235] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for acquiring video data from a monitoring device, a means for preprocessing the acquired video data, and a means for transmitting the preprocessed video data to an analysis device. This makes it possible to acquire video data in real time, detect abnormal behavior, and identify user emotions. The server also includes a means for anonymizing the analysis results and notifying the user terminal and related systems, and a means for storing the analysis results in a database and using them as learning data for the system. The server also includes a means for retraining the model of the analysis device based on new data to improve analysis accuracy. This makes it possible to continuously improve the analysis capabilities of the system while achieving both abnormal behavior detection accuracy and privacy protection.

[0236] A "surveillance device" is a device, such as a fixed camera or a drive recorder, that is installed to monitor a specific area.

[0237] "Video data" refers to visual information acquired from a surveillance device, typically consisting of a series of image frames.

[0238] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on the acquired video data to prepare it for analysis.

[0239] An "analysis device" is a device that uses advanced algorithms, such as deep learning models, to analyze pre-processed video data and detect abnormal behavior and emotions.

[0240] "Abnormal behavior" refers to movements or positions of people or objects that deviate from the normal movement of people or objects, and includes intrusions, violent acts, and abnormal stopping of vehicles.

[0241] The "emotion engine" is a system that analyzes a user's facial expressions and tone of voice from video data to identify emotions such as fear, surprise, and anger.

[0242] "Anonymization" refers to the process of removing personally identifiable information from analysis results to protect privacy. Specifically, it refers to processes such as blurring faces and masking audio.

[0243] A "user terminal" is a device, such as a smartphone or computer, that a user uses to receive information.

[0244] A "database" is a collection of data that stores analysis results and emotion analysis results and is used as learning data for the future.

[0245] "Model retraining" is the process of retraining the analysis device's algorithms based on new data to improve analysis accuracy.

[0246] The present invention relates to a system for acquiring video data from a monitoring device, analyzing the video data, and detecting abnormal behavior and emotions. Specific embodiments of the present invention will be described in detail below.

[0247] First, the server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. This video data is temporarily stored in storage on the server. The server then performs preprocessing on the acquired video data, performing noise removal and frame rate normalization. Software libraries such as OpenCV and FFmpeg are currently used for preprocessing, which are standard technologies.

[0248] The preprocessed video data is then sent to an analysis device, which uses machine learning libraries such as TensorFlow and PyTorch to analyze human movement and abnormal behavior using deep learning models. This analysis detects human movement and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0249] The analysis device also incorporates an emotion engine, which analyzes the user's facial expressions and tone of voice from the video data to identify emotions. Using facial expression recognition algorithms and voice analysis techniques, it is possible to determine the emotions expressed by the user (e.g., fear, surprise, anger, etc.). It is recommended to use the Facial-Emotion-Recognition library for this emotion identification.

[0250] Once the analysis is complete, the server retrieves the results and anonymizes them, specifically by blurring faces and masking audio to remove any personally identifiable information. This can be done with the help of dlib or other facial recognition libraries.

[0251] The server then notifies the appropriate user device and related systems of the anonymized analysis results and sentiment analysis results. Users receive notifications on their smartphones, PCs, or other devices and can take action as needed. If necessary, the server also securely shares the anonymized data with relevant agencies (police, security companies). This is done via a secure API.

[0252] The server also stores the analysis results and sentiment analysis results in a database. This stored data is also used as a dataset for future learning. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. A batch training method is used for retraining, and the model is updated.

[0253] As a concrete example, consider a scenario in which surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, an emotion engine analyzes the suspicious person's fear and tension from the video data and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0254] Through the above process, the system can efficiently detect abnormal behavior and respond quickly to deter crime, while also detecting potential dangers early through emotion analysis, further enhancing safety.

[0255] Example prompt sentence:

[0256] "Please explain the real-time suspicious person detection system for commercial facilities. Please include the names of the specific hardware and software, as well as the type of data processing that is performed."

[0257] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0258] Step 1:

[0259] The server receives video data in real time via the network from surveillance devices such as fixed cameras and dashcams. The input is the video data sent from the surveillance devices, and the output is the video data temporarily stored in the server's storage. This data is used in the next pre-processing step.

[0260] Step 2:

[0261] The server performs noise reduction and frame rate normalization on the acquired video data. The input is the video data saved in step 1, and the output is the preprocessed video data. The OpenCV library is used for noise reduction, and algorithms such as Gaussian Blur can be used. FFmpeg is used for frame rate normalization.

[0262] Step 3:

[0263] The server sends the preprocessed video data to the analysis device. The input is the video data preprocessed in step 2, and the output is the video data in the format required by the analysis device. This data transmission uses HTTP requests and socket communication.

[0264] Step 4:

[0265] The analysis device uses a deep learning model to analyze the preprocessed video data and detect abnormal behavior. The input is the video data sent in step 3, and the output is the analysis results of abnormal behavior (including timestamps and location information). Libraries such as TensorFlow and PyTorch are used for the analysis. Specifically, person detection algorithms such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector) are used.

[0266] Step 5:

[0267] The analysis device uses an emotion engine to analyze the user's emotions from the video data. The input is the same video data as in step 3, and the output is the emotion analysis results (emotional information such as fear, surprise, and anger). The Facial-Emotion-Recognition library is used for emotion analysis. Specific operations include a facial landmark detection algorithm and a facial expression classification algorithm.

[0268] Step 6:

[0269] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior analysis results and emotion analysis results obtained in steps 4 and 5, and the output is anonymized data. Specifically, the dlib library is used for face mosaic processing, and audio editing software such as Audacity is used for audio masking processing.

[0270] Step 7:

[0271] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized data from step 6, and the output is a notification message displayed on the user's device. This notification is sent using a push notification service such as Firebase Cloud Messaging or Apple Push Notification Service. Specifically, the server creates a notification message and sends it to the user's smartphone or PC.

[0272] Step 8:

[0273] The server stores the analysis results and sentiment analysis results in a database. The input is the analysis results and sentiment analysis results obtained in steps 4 and 5, and the output is the data stored in the database. This data will also be used as a training dataset in the future. Specifically, the server stores the data using an SQL database or NoSQL database.

[0274] Step 9:

[0275] The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. The input is the training data stored in the database, and the output is the retrained analysis model. Specific operations include batch training and updating the model weights. Possible libraries used are TensorFlow and PyTorch.

[0276] (Application example 2)

[0277] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0278] In modern society, security threats are increasing daily, and there is a need for rapid and accurate detection and response of abnormal behavior, especially in large commercial and public facilities. While numerous surveillance cameras are installed in such locations, it is difficult to monitor all camera footage in real time due to limited human resources. Furthermore, emotion analysis is also important in detecting abnormal behavior, making it necessary to discover and respond to potential threats early. Furthermore, there is a need for a method to solve these issues while protecting individual privacy.

[0279] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0280] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for the analysis device to analyze a person's emotions, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying a user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, and means for displaying the notified analysis results on a user terminal. This enables security staff to check anomalous behavior and emotion analysis results in real time and quickly recognize and respond to potential threats.

[0281] "Monitoring equipment" refers to equipment used to acquire video data, and primarily refers to fixed cameras and dashcams.

[0282] "Video data" refers to visual information acquired from a monitoring device, and is digital data that records movements and situations.

[0283] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on video data.

[0284] An "analysis device" is a device that uses preprocessed video data to analyze people's movements, abnormal behavior, and emotions.

[0285] "Abnormal behavior" refers to any unusual behavior or action, including, for example, intrusion, violent behavior, or abnormal stopping of a vehicle.

[0286] "Emotion analysis" is a technology that analyzes a person's facial expressions and tone of voice to identify specific emotions.

[0287] "Anonymization" is a process that removes personally identifiable information from the analysis results to protect the privacy of the data.

[0288] "User terminal" refers to electronic devices used to display analysis results, such as smartphones, PCs, and smart glasses.

[0289] "Related systems" refers to other systems or institutions that receive the analysis results, including police and security companies.

[0290] A "database" is a system for systematically storing and managing analysis results and learning data.

[0291] "Relearning" is a process of updating the model of the analysis device based on new data to improve the accuracy of the analysis.

[0292] "Notification" refers to the act of notifying the analysis results to the appropriate user terminal or related system.

[0293] The system according to the present invention can be implemented as a smart glasses application related to security services, and includes a monitoring device, a server, an analysis device, smart glasses, and related systems.

[0294] First, the server acquires video data in real time from monitoring devices (such as fixed cameras) connected to the network. This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to the analysis device.

[0295] The analyzer then receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). Furthermore, the analyzer incorporates an emotion engine that analyzes human facial expressions and speech tones in the video data to identify emotions. This emotion analysis can determine whether a person is displaying emotions such as fear, surprise, or anger.

[0296] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. Personally identifiable information is removed by methods such as facial mosaic and audio masking. Emotion analysis results are also anonymized. The server then notifies the appropriate user device and related systems of the analysis and emotion analysis results. Users receive notifications on their smart glasses and can check the results in real time. If necessary, the server also securely shares the anonymized data with relevant authorities (police, security companies).

[0297] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0298] As a concrete example, if security staff at a large shopping mall are patrolling wearing smart glasses, images from surveillance cameras will be displayed on the glasses in real time. At the same time, the system can detect intrusions and any abnormal behavior, and the emotions of individuals who exhibit suspicious behavior (such as fear or tension) can also be displayed on the glasses. This allows security staff to quickly recognize potential threats and respond efficiently.

[0299] Here is an example of a prompt to input to a generative AI model:

[0300] Run the following code to create an application that displays abnormal behavior and emotion analysis results in real time on smart glasses worn by security staff at a commercial facility.

[0301] The surveillance equipment video analysis system detects abnormal behavior and emotions in real time. The detected information is displayed on the security staff's smart glasses. Specifically, the system performs noise removal and frame rate normalization on the video data, abnormal behavior analysis using a deep learning model, and emotion analysis using an emotion engine, and displays the results on the smart glasses.

[0302] Libraries used:

[0303] cv2 (OpenCV)

[0304] Deep learning models (e.g., DeepFace)

[0305] Smart Glasses API Library

[0306] Required functions:

[0307] get_camera_feed(): Gets video data from the server

[0308] preprocess_frame(frame): Preprocesses video data

[0309] detect_anomalies(frame): Detect abnormal behavior

[0310] analyze_emotions(frame): Analyze emotions

[0311] display_on_Glass(TM)es(info): Display information on smart glasses

[0312] Please generate Python code that meets the above specifications.

[0313] As such, this system has a wide range of applications in the security field, particularly in terms of its ability to detect abnormal behavior in real time and respond quickly through emotion analysis.

[0314] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0315] Step 1:

[0316] The server acquires video data in real time from a monitoring device (such as a fixed camera). In this step, the server receives the video data sent from the monitoring device as streaming data and temporarily stores it in storage.

[0317] Step 2:

[0318] The server performs noise reduction and frame rate normalization on the acquired video data. The input is raw video data stored in storage, and the server applies noise reduction filters and frame rate unification processing to it before outputting the preprocessed video data.

[0319] Step 3:

[0320] The server transmits the preprocessed video data to the analysis device. The input is the preprocessed video data, and the server transmits this to the analysis device via the network.

[0321] Step 4:

[0322] The analysis device receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. The input of this step is the preprocessed video data, and the output is the detected abnormal behavior (e.g., intrusion, violent behavior, abnormal car stalling, etc.) along with its timestamp and location information.

[0323] Step 5:

[0324] The analyzer identifies emotions from video data by analyzing facial expressions and tone of speech. The input is pre-processed video data, and the output is the identified emotion (e.g., fear, surprise, anger, etc.). The analyzer does this using an emotion engine.

[0325] Step 6:

[0326] The server receives the analysis results and emotion analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior and emotion analysis results provided by the analysis device, and the output is the anonymized analysis results. The server performs operations such as face mosaic processing and audio masking.

[0327] Step 7:

[0328] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized analysis results, which the server notifies at the appropriate time. This allows the results to be displayed in real time on the user device (such as smart glasses).

[0329] Step 8:

[0330] The server stores the analysis results and sentiment analysis results in a database. The inputs to this step are the analysis results and sentiment analysis results, and the output is the stored data. The stored data can also be used as learning data in the future.

[0331] Step 9:

[0332] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is the newly saved data, which the server uses to retrain the deep learning model. The output is an analysis model with improved accuracy.

[0333] The detailed inputs, processing, and outputs for each step are shown below:

[0334] Step 1:

[0335] Input: Real-time video data from a surveillance device

[0336] Processing: The server receives the video data and temporarily stores it in storage.

[0337] Output: Raw video data stored in storage

[0338] Step 2:

[0339] Input: Raw video data stored in storage

[0340] Processing: Noise reduction filter applied, frame rate unified processing

[0341] Output: Pre-processed video data

[0342] Step 3:

[0343] Input: Preprocessed video data

[0344] Processing: Send data to an analysis device via a network

[0345] Output: Video data transmission to analysis device completed

[0346] Step 4:

[0347] Input: Preprocessed video data

[0348] Processing: Analyzing movement and abnormal behavior with deep learning models

[0349] Output: Detected abnormal behavior (timestamp and location information)

[0350] Step 5:

[0351] Input: Preprocessed video data

[0352] Processing: Identifying emotions with the emotion engine

[0353] Output: Identified emotion (e.g., fear, surprise, anger)

[0354] Step 6:

[0355] Input: Abnormal behavior and emotion analysis results

[0356] Processing: Anonymization by face mosaic processing, voice masking, etc.

[0357] Output: Anonymized analysis results

[0358] Step 7:

[0359] Input: Anonymized analysis results

[0360] Processing: Notify user terminals and related systems

[0361] Output: Notification to user terminal and related systems completed

[0362] Step 8:

[0363] Input: Analysis results and sentiment analysis results

[0364] Process: Save to database

[0365] Output: Data stored in the database

[0366] Step 9:

[0367] Input: Newly saved data

[0368] Processing: Retraining a deep learning model

[0369] Output: Improved analytical model

[0370] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0371] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0372] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0373] [Second embodiment]

[0374] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0375] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0376] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0377] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0378] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0379] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0380] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0381] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0382] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0383] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0384] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0385] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0386] A specific embodiment of a surveillance device video analysis system will be described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to prevent crime.

[0387] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0388] The analysis device then receives the preprocessed video data and analyzes it using deep learning models to detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0389] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, removing any personally identifiable information by pixelating faces and masking voices.

[0390] The server then notifies the appropriate user device and related systems of the analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take action. Additionally, the server can securely share anonymized data with relevant authorities (police, security companies) as needed.

[0391] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0392] Specific examples

[0393] Example of suspicious person detection in a commercial facility

[0394] Monitoring equipment installed in commercial facilities transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly.

[0395] In this way, surveillance video analysis systems can efficiently detect abnormal behavior and quickly respond to crime, thereby preventing crime. Furthermore, privacy protection is guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0396] The processing flow will be explained below.

[0397] Step 1:

[0398] The server acquires video data in real time from monitoring devices connected to the network. The monitoring devices come in a variety of forms, such as fixed cameras and dashcams, and the server connects to these devices to collect data.

[0399] Step 2:

[0400] The server temporarily stores the acquired video data in storage. This storage process is performed to back up the data so that it is not lost.

[0401] Step 3:

[0402] The server then begins pre-processing the video data stored in the storage, which includes noise removal, frame rate normalization, and resolution adjustment.

[0403] Step 4:

[0404] The server transmits the pre-processed video data to the analysis device, which transmits the data quickly and securely and ensures that the analysis device receives the data.

[0405] Step 5:

[0406] The analyzer receives the preprocessed data and uses deep learning models to analyze people's movements and abnormal behavior. In this process, deep learning algorithms detect abnormal behavior (such as intrusions or violent behavior) in the video.

[0407] Step 6:

[0408] When the analysis device detects abnormal behavior, it records the timestamp and location information and sends the analysis results to the server.

[0409] Step 7:

[0410] The server receives the analysis results, anonymizes any information that could identify individuals to protect privacy, and applies mosaic processing and audio masking to the video data.

[0411] Step 8:

[0412] The server notifies the user and related systems of the anonymized analysis results, and the user receives the notification on their smartphone or PC and can take action.

[0413] Step 9:

[0414] Users can check the notification and take immediate action if necessary. For example, a commercial facility manager can work with security staff to investigate the situation and take appropriate action.

[0415] Step 10:

[0416] The server stores the analysis results in a database, which will be used as learning data in the future.

[0417] Step 11:

[0418] The server periodically uses the stored data to retrain the analysis device's model, improving analysis accuracy and enabling more effective detection of abnormal behavior.

[0419] Step 12:

[0420] The server applies the retrained model to the system and continues analyzing the monitoring device data with the new analytical accuracy.

[0421] Example 1

[0422] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0423] Conventional surveillance systems have a problem in that they have low accuracy in analyzing video data, making it difficult to accurately detect abnormal behavior. Furthermore, they lack the functionality to detect and notify abnormal behavior in real time while protecting individual privacy, which often results in delayed response. To solve these problems, an efficient and highly accurate video data analysis system is required.

[0424] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0425] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for performing noise reduction and frame rate normalization on the preprocessed video data, means for recording timestamps and location information of anomalous behavior detected by the analysis device, means for performing face mosaic processing and audio masking when acquiring the analysis results and anonymizing the data for privacy protection, means for notifying the user terminal of the analysis results using an API or push notification system, and means for acquiring video data from the monitoring device in real time, preprocessing the video data, and transferring the video data to the analysis device, thereby enabling highly accurate detection of anomalous behavior.

[0426] A "monitoring device" is a device that continuously captures images of a monitored area and generates video data, and specifically includes fixed cameras and drive recorders.

[0427] "Video data" refers to video information acquired by a monitoring device and is made up of successive frames of images.

[0428] "Preprocessing" refers to processing such as noise removal and frame rate normalization before analyzing video data.

[0429] The "analysis device" is a device that runs a deep learning model to detect abnormal behavior using preprocessed video data as input.

[0430] "Abnormal behavior" refers to any unusual behavior or condition that occurs within the monitored area, including, for example, intrusion, violent behavior, or abnormal vehicle stalls.

[0431] "Anonymization" is a process carried out to protect information that could identify an individual from the analysis results, and specifically includes blurring faces and masking audio.

[0432] A "user terminal" is an electronic device capable of receiving and displaying analysis results, such as a smartphone or a personal computer.

[0433] A "database" is a system for efficiently storing and managing analysis results and learning data.

[0434] "Retraining" is the process of updating a deep learning model based on new data to improve analysis accuracy.

[0435] "Noise reduction" is the process of removing unnecessary background information and noise from video data.

[0436] "Frame rate normalization" is a process of standardizing the frame rate of video data to a fixed value.

[0437] A "timestamp" is information indicating a specific time, and is recorded in video data or analysis results.

[0438] "Location information" is information indicating the location where abnormal behavior was detected, and includes geographic coordinates, addresses, etc.

[0439] "API" stands for Application Program Interface, an interface that allows software functions and data to be used by other programs.

[0440] A "push notification system" is a system that sends notifications from a server to a user terminal in real time.

[0441] "Face mosaic processing" is a process of processing video data so that an individual's face cannot be identified, and includes polygon and blur effects.

[0442] "Audio masking" is a process of processing audio data to prevent individuals from being identified.

[0443] This invention describes a specific embodiment of a surveillance device video analysis system. This system is mainly composed of a server, a terminal, and a user, and aims to efficiently analyze video data acquired from surveillance devices and prevent crime.

[0444] First, the server acquires video data in real time from monitoring devices connected to the network (for example, fixed cameras or drive recorders). The server temporarily stores this video data in storage. To acquire the video data, for example, the video stream URL of an IP camera can be used.

[0445] The server then performs preprocessing on the acquired video data. This preprocessing includes noise reduction and frame rate normalization. For example, noise reduction using a Gaussian filter and standardizing the video frame rate to 30 fps are performed. This preprocessing allows for the efficient execution of subsequent analysis processes.

[0446] The server then sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the data and detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for abnormal behavior.

[0447] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, specifically by blurring faces (for example, using AWS Rekognition) and masking voices to remove personally identifiable information.

[0448] The server then notifies the user's device (e.g., smartphone or PC) and related systems of the analysis results. Notifications are sent using APIs or push notification systems. Users can receive notifications on their devices and take action. If necessary, the server also securely shares anonymized data with related organizations (e.g., police or security companies).

[0449] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device model based on new data to improve analysis accuracy. For example, TensorFlow or PyTorch can be used for this retraining.

[0450] Specific examples

[0451] Example of suspicious person detection in a commercial facility

[0452] Surveillance devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly. In this way, the surveillance device video analysis system efficiently detects abnormal behavior and acts quickly to deter crime. Privacy protection is also guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0453] Example prompts for generative AI models

[0454] text

[0455] Surveillance cameras installed in a commercial facility send video footage to a server in real time. The server preprocesses this video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the smartphone of the commercial facility manager. The manager receives the alert and responds promptly. Please explain this series of processes in easy-to-understand text.

[0456] This system makes it possible to efficiently analyze video data acquired from surveillance cameras, quickly detect and notify users of abnormal behavior, and provides a sense of security to users as it also takes privacy protection into consideration.

[0457] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0458] Step 1:

[0459] Acquiring and storing video data

[0460] The server acquires video data in real time from surveillance devices (e.g., fixed cameras or dashcams). This is done using the IP address or stream URL of the surveillance camera. The server then temporarily stores the acquired video data in storage. Specifically, it establishes a streaming connection for the video data and stores the acquired video frames sequentially.

[0461] Input: Surveillance video stream

[0462] Output: Temporarily saved video data

[0463] Step 2:

[0464] Video data preprocessing

[0465] The server performs preprocessing on the stored video data. This preprocessing includes noise reduction and frame rate normalization. A Gaussian filter is used for noise reduction, and frame rate normalization involves thinning and interpolating frames along the time axis. This improves data quality and increases the accuracy of analysis.

[0466] Input: Temporarily stored video data

[0467] Output: Pre-processed video data

[0468] Step 3:

[0469] Analysis of preprocessed video data

[0470] The server sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the video data and detect human movements and abnormal behavior. This analysis model operates in real time using pre-trained parameters.

[0471] Input: Preprocessed video data

[0472] Output: Analysis results (detection results of abnormal behavior, timestamp, location information)

[0473] Step 4:

[0474] Anonymization of analysis results

[0475] The server receives the analysis results and performs face blurring and audio masking to protect privacy. Face blurring is performed using image recognition software (e.g., AWS Rekognition), and audio masking is performed using technology that removes specific frequency bands.

[0476] Input: Analysis results

[0477] Output: Anonymized analysis results (face mosaic and voice masked data)

[0478] Step 5:

[0479] Notification to user devices and related systems

[0480] The server notifies the user's device and related systems of the anonymized analysis results using an API or push notification system. Users can receive the notifications on their smartphones or computers and respond promptly.

[0481] Input: Anonymized analysis results

[0482] Output: Notification to user device (push notification to smartphone or PC)

[0483] Step 6:

[0484] Saving analysis results and retraining models

[0485] The server stores the analysis results in a database and uses them as training data for the system. The analysis device model is periodically retrained based on new data to improve analysis accuracy. This retraining process uses TensorFlow and PyTorch to update the model parameters.

[0486] Input: Analysis results

[0487] Output: Updated analytical model, saved training data

[0488] These steps enable the system to achieve highly accurate detection of abnormal behavior, protect privacy, and support rapid response.

[0489] (Application example 1)

[0490] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0491] Conventional monitoring devices and systems often have delays in detecting abnormal behavior and subsequent responses, making it difficult to respond quickly. Another issue is that data privacy protection is insufficient, increasing the risk of personal information leaks. Furthermore, improving system training and analysis accuracy takes a lot of time and effort. There is a need for a system that can solve these issues and detect and respond to abnormal behavior more efficiently and quickly.

[0492] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0493] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect abnormal behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for notifying a smartphone application of abnormal behavior alerts in real time, and means for the user to view past abnormal behavior detection history. This enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases user peace of mind.

[0494] A "surveillance device" is a device for acquiring video data, such as a fixed camera or a drive recorder.

[0495] "Video data" refers to visual information acquired from a monitoring device, and is data used to detect abnormal behavior.

[0496] "Preprocessing" refers to the process of removing noise from the acquired video data, normalizing the frame rate, and otherwise processing the data to make it easier to analyze.

[0497] The "analysis device" is a device for analyzing pre-processed video data and detecting abnormal behavior.

[0498] "Abnormal behavior" refers to behavior that is different from the norm, such as the intrusion of suspicious individuals, violent behavior, or abnormal stopping of vehicles.

[0499] "Anonymization" is a process that removes information that can identify individuals from the analysis results in order to protect privacy.

[0500] A "user terminal" is a device, such as a smartphone or PC, that a user uses to check analysis results and receive notifications.

[0501] "Related systems" are external systems, such as police and security companies, that are necessary for sharing analysis results.

[0502] A "database" is a data storage device that stores analysis results and uses them as learning data for the system.

[0503] "Relearning" is the process of retraining the analysis device's model based on new data to improve analysis accuracy.

[0504] A "smartphone application" is software that runs on a smartphone and alerts users of abnormal behavior in real time.

[0505] An "abnormal behavior alert" is a warning or information that quickly notifies the user of detected abnormal behavior.

[0506] "History" is a record of abnormal behavior detected in the past, and is data including information that can be confirmed by the user.

[0507] A specific embodiment of a surveillance device video analysis system is described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from a surveillance device, detects abnormal behavior, and promptly notifies the user.

[0508] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored on storage (e.g., SSD or HDD). The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. For this purpose, libraries such as OpenCV are used. The processed video data is then sent to an analysis device.

[0509] The analysis device then receives the preprocessed video data and analyzes it using deep learning models (e.g., Keras or TensorFlow). This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0510] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. This includes blurring faces and masking audio to remove any personally identifiable information. The server then notifies the appropriate user device (e.g., smartphone, PC) and related systems (e.g., police, security companies) of the analysis results. Users receive alerts of abnormal behavior in real time through a smartphone application, allowing them to respond quickly. Users can also view past abnormal behavior detection history and consider countermeasures.

[0511] Furthermore, the server stores the analysis results in a database, which is used as training data for future systems. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. Retraining is performed using deep learning libraries such as Keras and TensorFlow.

[0512] As a specific example, consider a case where surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond quickly. In this way, the surveillance equipment video analysis system efficiently detects abnormal behavior and aims to deter crime by responding quickly.

[0513] Analysis using generative AI models uses prompts like the following:

[0514] "Analyze this video data to detect any abnormal behavior (e.g., unauthorized entry, violent behavior, or abnormal stopping). Include timestamps, detailed information, and location information in the detection results. Anonymize the data to protect privacy."

[0515] The above is a specific embodiment. This system enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases the sense of security of users.

[0516] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0517] Step 1:

[0518] The server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. The input is video data from the monitoring devices, and the output is data that is temporarily stored in storage. Specifically, the server receives video data sent from the monitoring devices and temporarily stores it.

[0519] Step 2:

[0520] The server performs preprocessing on the acquired video data, such as noise removal and frame rate normalization. The input here is the video data stored in storage, and the output is preprocessed, clean video data. Specifically, the server uses the OpenCV library to remove noise from the video and adjust the frame rate to a constant value.

[0521] Step 3:

[0522] The server sends the preprocessed video data to the analysis device. The input is the preprocessed video data, and the output is the data sent to the analysis device. Specifically, the preprocessed data is transferred to the analysis device via a network.

[0523] Step 4:

[0524] The analysis device receives the preprocessed video data and analyzes it using a deep learning model. This analysis detects abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The input is the preprocessed video data sent to the analysis device, and the output is the abnormal behavior detection results (timestamp, detailed information, and location information). Specifically, the data is analyzed using a generative AI model using Keras and TensorFlow.

[0525] Step 5:

[0526] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior detection results from the analysis device, and the output is anonymized data. Specific operations include face mosaic processing and voice masking to remove personally identifiable information.

[0527] Step 6:

[0528] The server notifies the user device (smartphone, PC) and related systems (police, security companies) of the anonymized analysis results. The input is the anonymized analysis results, and the output is the notified data. Specifically, the server sends the data to the user device or external system via an HTTP request, etc.

[0529] Step 7:

[0530] Users can receive alerts of abnormal behavior in real time through a smartphone application. The input is notification data from the server, and the output is alert information displayed on the smartphone application. Specifically, the application notification function receives the data and displays it on the screen.

[0531] Step 8:

[0532] Users can view the history of past abnormal behavior detections using a smartphone application. The input is the analysis results stored in the database, and the output is the history information displayed on the application. Specifically, the application queries the database, retrieves past detection results, and displays them.

[0533] Step 9:

[0534] The server stores the analysis results in a database and uses them as learning data for the system. The input is the analysis results, and the output is the data stored in the database. Specifically, the analysis results are formalized and stored in the database.

[0535] Step 10:

[0536] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is new data and past analysis results, and the output is the retrained model. Specifically, the model is updated using the retraining functions of Keras or TensorFlow.

[0537] This series of steps enables efficient and rapid detection and response to abnormal behavior, ensuring the safety of users.

[0538] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0539] Below, we will explain a specific embodiment that combines a surveillance video analysis system with an emotion engine. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to deter crime and recognize user emotions.

[0540] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0541] The analysis device then receives the preprocessed video data and uses deep learning models to analyze human movements and abnormal behaviors (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behaviors.

[0542] Furthermore, the analysis device is equipped with an emotion engine that analyzes the user's facial expressions and tone of voice from the video data to identify their emotions. This emotion analysis can determine whether the user is expressing emotions such as fear, surprise, or anger.

[0543] Once the analysis is complete, the server retrieves the results and anonymizes them to protect privacy. It uses techniques like face blurring and voice masking to remove any personally identifiable information. Sentiment analysis results are also anonymized.

[0544] The server then notifies the appropriate user device and related systems of the analysis and sentiment analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take appropriate action. Additionally, the server can securely share anonymized data with relevant agencies (police, security companies) as needed.

[0545] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0546] Specific examples

[0547] Example of suspicious person detection in a commercial facility

[0548] Monitoring devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, the emotion engine analyzes the video data to determine the suspicious person's fear or tension, and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0549] In this way, a system that combines a surveillance video analysis system and an emotion engine can efficiently detect abnormal behavior and quickly respond to it, thereby preventing crime. Emotion analysis also enables early detection of potential danger, further enhancing safety. Privacy protection is also guaranteed, improving users' sense of security. This system can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0550] The processing flow will be explained below.

[0551] Step 1:

[0552] The server receives video data in real time from network-connected surveillance devices, such as fixed cameras and dashcams, and connects to these devices to receive their signals.

[0553] Step 2:

[0554] The server temporarily stores the acquired video data in storage so that the data can be referenced later for processing.

[0555] Step 3:

[0556] The server performs preprocessing on the video data stored in the storage, including noise removal, frame rate normalization, and resolution adjustment.

[0557] Step 4:

[0558] The server transmits the pre-processed video data to the analysis equipment, and this transmission is fast and secure, ensuring that the data reaches the analysis equipment reliably.

[0559] Step 5:

[0560] The analysis device receives the preprocessed video data and uses deep learning models to analyze people and their behavior in the video, detecting abnormal behavior (such as intrusions, violent behavior, or abnormal vehicle stalls).

[0561] Step 6:

[0562] If the analysis device detects abnormal behavior, it records the timestamp and location information and sends this to the server as the analysis result.

[0563] Step 7:

[0564] The emotion engine included in the analysis device analyzes the user's facial expressions and tone of voice from the video data to generate emotion data, which can include fear, surprise, anger, etc.

[0565] Step 8:

[0566] The analysis device sends the emotion data to the server, where it is treated in the same way as the analysis results of abnormal behavior.

[0567] Step 9:

[0568] The server receives the analysis results and emotion data, and anonymizes them to protect privacy. It also applies mosaic processing to the video and masks the audio data.

[0569] Step 10:

[0570] The server notifies the user device and related systems of the anonymized analysis results and emotion data, and if notification is required, sends an alert to specific users or systems.

[0571] Step 11:

[0572] Users receive notifications on their smartphones or PCs and can take action as necessary. For example, a commercial facility manager can immediately contact security staff to check the situation and take appropriate measures.

[0573] Step 12:

[0574] The server stores the analysis results and emotion data in a database, which will be used as learning data for the system in the future.

[0575] Step 13:

[0576] The server periodically uses the saved data to retrain the analysis device's model, improving the model's analysis accuracy and enabling it to detect abnormal behavior and user emotions with greater accuracy.

[0577] Step 14:

[0578] The server applies the retrained model to the system and continues analyzing the video data from the surveillance device with improved analysis accuracy.

[0579] Example 2

[0580] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0581] Conventional surveillance systems were specialized in detecting abnormal behavior, but lacked the ability to analyze people's emotions, making it difficult to detect potential dangers early on. Furthermore, from the perspective of protecting privacy, it was necessary to adequately protect personally identifiable information.

[0582] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for acquiring video data from a monitoring device, a means for preprocessing the acquired video data, and a means for transmitting the preprocessed video data to an analysis device. This makes it possible to acquire video data in real time, detect abnormal behavior, and identify user emotions. The server also includes a means for anonymizing the analysis results and notifying the user terminal and related systems, and a means for storing the analysis results in a database and using them as learning data for the system. The server also includes a means for retraining the model of the analysis device based on new data to improve analysis accuracy. This makes it possible to continuously improve the analysis capabilities of the system while achieving both abnormal behavior detection accuracy and privacy protection.

[0583] A "surveillance device" is a device, such as a fixed camera or a drive recorder, that is installed to monitor a specific area.

[0584] "Video data" refers to visual information acquired from a surveillance device, typically consisting of a series of image frames.

[0585] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on the acquired video data to prepare it for analysis.

[0586] An "analysis device" is a device that uses advanced algorithms, such as deep learning models, to analyze pre-processed video data and detect abnormal behavior and emotions.

[0587] "Abnormal behavior" refers to movements or positions of people or objects that deviate from the normal movement of people or objects, and includes intrusions, violent acts, and abnormal stopping of vehicles.

[0588] The "emotion engine" is a system that analyzes a user's facial expressions and tone of voice from video data to identify emotions such as fear, surprise, and anger.

[0589] "Anonymization" refers to the process of removing personally identifiable information from analysis results to protect privacy. Specifically, it refers to processes such as blurring faces and masking audio.

[0590] A "user terminal" is a device, such as a smartphone or computer, that a user uses to receive information.

[0591] A "database" is a collection of data that stores analysis results and emotion analysis results and is used as learning data for the future.

[0592] "Model retraining" is the process of retraining the analysis device's algorithms based on new data to improve analysis accuracy.

[0593] The present invention relates to a system for acquiring video data from a monitoring device, analyzing the video data, and detecting abnormal behavior and emotions. Specific embodiments of the present invention will be described in detail below.

[0594] First, the server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. This video data is temporarily stored in storage on the server. The server then performs preprocessing on the acquired video data, performing noise removal and frame rate normalization. Software libraries such as OpenCV and FFmpeg are currently used for preprocessing, which are standard technologies.

[0595] The preprocessed video data is then sent to an analysis device, which uses machine learning libraries such as TensorFlow and PyTorch to analyze human movement and abnormal behavior using deep learning models. This analysis detects human movement and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0596] The analysis device also incorporates an emotion engine, which analyzes the user's facial expressions and tone of voice from the video data to identify emotions. Using facial expression recognition algorithms and voice analysis techniques, it is possible to determine the emotions expressed by the user (e.g., fear, surprise, anger, etc.). It is recommended to use the Facial-Emotion-Recognition library for this emotion identification.

[0597] Once the analysis is complete, the server retrieves the results and anonymizes them, specifically by blurring faces and masking audio to remove any personally identifiable information. This can be done with the help of dlib or other facial recognition libraries.

[0598] The server then notifies the appropriate user device and related systems of the anonymized analysis results and sentiment analysis results. Users receive notifications on their smartphones, PCs, or other devices and can take action as needed. If necessary, the server also securely shares the anonymized data with relevant agencies (police, security companies). This is done via a secure API.

[0599] The server also stores the analysis results and sentiment analysis results in a database. This stored data is also used as a dataset for future learning. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. A batch training method is used for retraining, and the model is updated.

[0600] As a concrete example, consider a scenario in which surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, an emotion engine analyzes the suspicious person's fear and tension from the video data and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0601] Through the above process, the system can efficiently detect abnormal behavior and respond quickly to deter crime, while also detecting potential dangers early through emotion analysis, further enhancing safety.

[0602] Example prompt sentence:

[0603] "Please explain the real-time suspicious person detection system for commercial facilities. Please include the names of the specific hardware and software, as well as the type of data processing that is performed."

[0604] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0605] Step 1:

[0606] The server receives video data in real time via the network from surveillance devices such as fixed cameras and dashcams. The input is the video data sent from the surveillance devices, and the output is the video data temporarily stored in the server's storage. This data is used in the next pre-processing step.

[0607] Step 2:

[0608] The server performs noise reduction and frame rate normalization on the acquired video data. The input is the video data saved in step 1, and the output is the preprocessed video data. The OpenCV library is used for noise reduction, and algorithms such as Gaussian Blur can be used. FFmpeg is used for frame rate normalization.

[0609] Step 3:

[0610] The server sends the preprocessed video data to the analysis device. The input is the video data preprocessed in step 2, and the output is the video data in the format required by the analysis device. This data transmission uses HTTP requests and socket communication.

[0611] Step 4:

[0612] The analysis device uses a deep learning model to analyze the preprocessed video data and detect abnormal behavior. The input is the video data sent in step 3, and the output is the analysis results of abnormal behavior (including timestamps and location information). Libraries such as TensorFlow and PyTorch are used for the analysis. Specifically, person detection algorithms such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector) are used.

[0613] Step 5:

[0614] The analysis device uses an emotion engine to analyze the user's emotions from the video data. The input is the same video data as in step 3, and the output is the emotion analysis results (emotional information such as fear, surprise, and anger). The Facial-Emotion-Recognition library is used for emotion analysis. Specific operations include a facial landmark detection algorithm and a facial expression classification algorithm.

[0615] Step 6:

[0616] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior analysis results and emotion analysis results obtained in steps 4 and 5, and the output is anonymized data. Specifically, the dlib library is used for face mosaic processing, and audio editing software such as Audacity is used for audio masking processing.

[0617] Step 7:

[0618] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized data from step 6, and the output is a notification message displayed on the user's device. This notification is sent using a push notification service such as Firebase Cloud Messaging or Apple Push Notification Service. Specifically, the server creates a notification message and sends it to the user's smartphone or PC.

[0619] Step 8:

[0620] The server stores the analysis results and sentiment analysis results in a database. The input is the analysis results and sentiment analysis results obtained in steps 4 and 5, and the output is the data stored in the database. This data will also be used as a training dataset in the future. Specifically, the server stores the data using an SQL database or NoSQL database.

[0621] Step 9:

[0622] The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. The input is the training data stored in the database, and the output is the retrained analysis model. Specific operations include batch training and updating the model weights. Possible libraries used are TensorFlow and PyTorch.

[0623] (Application example 2)

[0624] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0625] In modern society, security threats are increasing daily, and there is a need for rapid and accurate detection and response of abnormal behavior, especially in large commercial and public facilities. While numerous surveillance cameras are installed in such locations, it is difficult to monitor all camera footage in real time due to limited human resources. Furthermore, emotion analysis is also important in detecting abnormal behavior, making it necessary to discover and respond to potential threats early. Furthermore, there is a need for a method to solve these issues while protecting individual privacy.

[0626] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0627] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for the analysis device to analyze a person's emotions, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying a user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, and means for displaying the notified analysis results on a user terminal. This enables security staff to check anomalous behavior and emotion analysis results in real time and quickly recognize and respond to potential threats.

[0628] "Monitoring equipment" refers to equipment used to acquire video data, and primarily refers to fixed cameras and dashcams.

[0629] "Video data" refers to visual information acquired from a monitoring device, and is digital data that records movements and situations.

[0630] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on video data.

[0631] An "analysis device" is a device that uses preprocessed video data to analyze people's movements, abnormal behavior, and emotions.

[0632] "Abnormal behavior" refers to any unusual behavior or action, including, for example, intrusion, violent behavior, or abnormal stopping of a vehicle.

[0633] "Emotion analysis" is a technology that analyzes a person's facial expressions and tone of voice to identify specific emotions.

[0634] "Anonymization" is a process that removes personally identifiable information from the analysis results to protect the privacy of the data.

[0635] "User terminal" refers to electronic devices used to display analysis results, such as smartphones, PCs, and smart glasses.

[0636] "Related systems" refers to other systems or institutions that receive the analysis results, including police and security companies.

[0637] A "database" is a system for systematically storing and managing analysis results and learning data.

[0638] "Relearning" is a process of updating the model of the analysis device based on new data to improve the accuracy of the analysis.

[0639] "Notification" refers to the act of notifying the analysis results to the appropriate user terminal or related system.

[0640] The system according to the present invention can be implemented as a smart glasses application related to security services, and includes a monitoring device, a server, an analysis device, smart glasses, and related systems.

[0641] First, the server acquires video data in real time from monitoring devices (such as fixed cameras) connected to the network. This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to the analysis device.

[0642] The analyzer then receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). Furthermore, the analyzer incorporates an emotion engine that analyzes human facial expressions and speech tones in the video data to identify emotions. This emotion analysis can determine whether a person is displaying emotions such as fear, surprise, or anger.

[0643] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. Personally identifiable information is removed by methods such as facial mosaic and audio masking. Emotion analysis results are also anonymized. The server then notifies the appropriate user device and related systems of the analysis and emotion analysis results. Users receive notifications on their smart glasses and can check the results in real time. If necessary, the server also securely shares the anonymized data with relevant authorities (police, security companies).

[0644] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0645] As a concrete example, if security staff at a large shopping mall are patrolling wearing smart glasses, images from surveillance cameras will be displayed on the glasses in real time. At the same time, the system can detect intrusions and any abnormal behavior, and the emotions of individuals who exhibit suspicious behavior (such as fear or tension) can also be displayed on the glasses. This allows security staff to quickly recognize potential threats and respond efficiently.

[0646] Here is an example of a prompt to input to a generative AI model:

[0647] Run the following code to create an application that displays abnormal behavior and emotion analysis results in real time on smart glasses worn by security staff at a commercial facility.

[0648] The surveillance equipment video analysis system detects abnormal behavior and emotions in real time. The detected information is displayed on the security staff's smart glasses. Specifically, the system performs noise removal and frame rate normalization on the video data, abnormal behavior analysis using a deep learning model, and emotion analysis using an emotion engine, and displays the results on the smart glasses.

[0649] Libraries used:

[0650] cv2 (OpenCV)

[0651] Deep learning models (e.g., DeepFace)

[0652] Smart Glasses API Library

[0653] Required functions:

[0654] get_camera_feed(): Gets video data from the server

[0655] preprocess_frame(frame): Preprocesses video data

[0656] detect_anomalies(frame): Detect abnormal behavior

[0657] analyze_emotions(frame): Analyze emotions

[0658] display_on_glasses(info): Display information on smart glasses

[0659] Please generate Python code that meets the above specifications.

[0660] As such, this system has a wide range of applications in the security field, particularly in terms of its ability to detect abnormal behavior in real time and respond quickly through emotion analysis.

[0661] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0662] Step 1:

[0663] The server acquires video data in real time from a monitoring device (such as a fixed camera). In this step, the server receives the video data sent from the monitoring device as streaming data and temporarily stores it in storage.

[0664] Step 2:

[0665] The server performs noise reduction and frame rate normalization on the acquired video data. The input is raw video data stored in storage, and the server applies noise reduction filters and frame rate unification processing to it before outputting the preprocessed video data.

[0666] Step 3:

[0667] The server transmits the preprocessed video data to the analysis device. The input is the preprocessed video data, and the server transmits this to the analysis device via the network.

[0668] Step 4:

[0669] The analysis device receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. The input of this step is the preprocessed video data, and the output is the detected abnormal behavior (e.g., intrusion, violent behavior, abnormal car stalling, etc.) along with its timestamp and location information.

[0670] Step 5:

[0671] The analyzer identifies emotions from video data by analyzing facial expressions and tone of speech. The input is pre-processed video data, and the output is the identified emotion (e.g., fear, surprise, anger, etc.). The analyzer does this using an emotion engine.

[0672] Step 6:

[0673] The server receives the analysis results and emotion analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior and emotion analysis results provided by the analysis device, and the output is the anonymized analysis results. The server performs operations such as face mosaic processing and audio masking.

[0674] Step 7:

[0675] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized analysis results, which the server notifies at the appropriate time. This allows the results to be displayed in real time on the user device (such as smart glasses).

[0676] Step 8:

[0677] The server stores the analysis results and sentiment analysis results in a database. The inputs to this step are the analysis results and sentiment analysis results, and the output is the stored data. The stored data can also be used as learning data in the future.

[0678] Step 9:

[0679] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is the newly saved data, which the server uses to retrain the deep learning model. The output is an analysis model with improved accuracy.

[0680] The detailed inputs, processing, and outputs for each step are shown below:

[0681] Step 1:

[0682] Input: Real-time video data from a surveillance device

[0683] Processing: The server receives the video data and temporarily stores it in storage.

[0684] Output: Raw video data stored in storage

[0685] Step 2:

[0686] Input: Raw video data stored in storage

[0687] Processing: Noise reduction filter applied, frame rate unified processing

[0688] Output: Pre-processed video data

[0689] Step 3:

[0690] Input: Preprocessed video data

[0691] Processing: Send data to an analysis device via a network

[0692] Output: Video data transmission to analysis device completed

[0693] Step 4:

[0694] Input: Preprocessed video data

[0695] Processing: Analyzing movement and abnormal behavior with deep learning models

[0696] Output: Detected abnormal behavior (timestamp and location information)

[0697] Step 5:

[0698] Input: Preprocessed video data

[0699] Processing: Identifying emotions with the emotion engine

[0700] Output: Identified emotion (e.g., fear, surprise, anger)

[0701] Step 6:

[0702] Input: Abnormal behavior and emotion analysis results

[0703] Processing: Anonymization by face mosaic processing, voice masking, etc.

[0704] Output: Anonymized analysis results

[0705] Step 7:

[0706] Input: Anonymized analysis results

[0707] Processing: Notify user terminals and related systems

[0708] Output: Notification to user terminal and related systems completed

[0709] Step 8:

[0710] Input: Analysis results and sentiment analysis results

[0711] Process: Save to database

[0712] Output: Data stored in the database

[0713] Step 9:

[0714] Input: Newly saved data

[0715] Processing: Retraining a deep learning model

[0716] Output: Improved analytical model

[0717] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0718] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0719] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0720] [Third embodiment]

[0721] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0722] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0723] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0724] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0725] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0726] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0727] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0728] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0729] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0730] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0731] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0732] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0733] A specific embodiment of a surveillance device video analysis system will be described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to prevent crime.

[0734] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0735] The analysis device then receives the preprocessed video data and analyzes it using deep learning models to detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0736] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, removing any personally identifiable information by pixelating faces and masking voices.

[0737] The server then notifies the appropriate user device and related systems of the analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take action. Additionally, the server can securely share anonymized data with relevant authorities (police, security companies) as needed.

[0738] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0739] Specific examples

[0740] Example of suspicious person detection in a commercial facility

[0741] Monitoring equipment installed in commercial facilities transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly.

[0742] In this way, surveillance video analysis systems can efficiently detect abnormal behavior and quickly respond to crime, thereby preventing crime. Furthermore, privacy protection is guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0743] The processing flow will be explained below.

[0744] Step 1:

[0745] The server acquires video data in real time from monitoring devices connected to the network. The monitoring devices come in a variety of forms, such as fixed cameras and dashcams, and the server connects to these devices to collect data.

[0746] Step 2:

[0747] The server temporarily stores the acquired video data in storage. This storage process is performed to back up the data so that it is not lost.

[0748] Step 3:

[0749] The server then begins pre-processing the video data stored in the storage, which includes noise removal, frame rate normalization, and resolution adjustment.

[0750] Step 4:

[0751] The server transmits the pre-processed video data to the analysis device, which transmits the data quickly and securely and ensures that the analysis device receives the data.

[0752] Step 5:

[0753] The analyzer receives the preprocessed data and uses deep learning models to analyze people's movements and abnormal behavior. In this process, deep learning algorithms detect abnormal behavior (such as intrusions or violent behavior) in the video.

[0754] Step 6:

[0755] When the analysis device detects abnormal behavior, it records the timestamp and location information and sends the analysis results to the server.

[0756] Step 7:

[0757] The server receives the analysis results, anonymizes any information that could identify individuals to protect privacy, and applies mosaic processing and audio masking to the video data.

[0758] Step 8:

[0759] The server notifies the user and related systems of the anonymized analysis results, and the user receives the notification on their smartphone or PC and can take action.

[0760] Step 9:

[0761] Users can check the notification and take immediate action if necessary. For example, a commercial facility manager can work with security staff to investigate the situation and take appropriate action.

[0762] Step 10:

[0763] The server stores the analysis results in a database, which will be used as learning data in the future.

[0764] Step 11:

[0765] The server periodically uses the stored data to retrain the analysis device's model, improving analysis accuracy and enabling more effective detection of abnormal behavior.

[0766] Step 12:

[0767] The server applies the retrained model to the system and continues analyzing the monitoring device data with the new analytical accuracy.

[0768] Example 1

[0769] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0770] Conventional surveillance systems have a problem in that they have low accuracy in analyzing video data, making it difficult to accurately detect abnormal behavior. Furthermore, they lack the functionality to detect and notify abnormal behavior in real time while protecting individual privacy, which often results in delayed response. To solve these problems, an efficient and highly accurate video data analysis system is required.

[0771] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0772] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for performing noise reduction and frame rate normalization on the preprocessed video data, means for recording timestamps and location information of anomalous behavior detected by the analysis device, means for performing face mosaic processing and audio masking when acquiring the analysis results and anonymizing the data for privacy protection, means for notifying the user terminal of the analysis results using an API or push notification system, and means for acquiring video data from the monitoring device in real time, preprocessing the video data, and transferring the video data to the analysis device, thereby enabling highly accurate detection of anomalous behavior.

[0773] A "monitoring device" is a device that continuously captures images of a monitored area and generates video data, and specifically includes fixed cameras and drive recorders.

[0774] "Video data" refers to video information acquired by a monitoring device and is made up of successive frames of images.

[0775] "Preprocessing" refers to processing such as noise removal and frame rate normalization before analyzing video data.

[0776] The "analysis device" is a device that runs a deep learning model to detect abnormal behavior using preprocessed video data as input.

[0777] "Abnormal behavior" refers to any unusual behavior or condition that occurs within the monitored area, including, for example, intrusion, violent behavior, or abnormal vehicle stalls.

[0778] "Anonymization" is a process carried out to protect information that could identify an individual from the analysis results, and specifically includes blurring faces and masking audio.

[0779] A "user terminal" is an electronic device capable of receiving and displaying analysis results, such as a smartphone or a personal computer.

[0780] A "database" is a system for efficiently storing and managing analysis results and learning data.

[0781] "Retraining" is the process of updating a deep learning model based on new data to improve analysis accuracy.

[0782] "Noise reduction" is the process of removing unnecessary background information and noise from video data.

[0783] "Frame rate normalization" is a process of standardizing the frame rate of video data to a fixed value.

[0784] A "timestamp" is information indicating a specific time, and is recorded in video data or analysis results.

[0785] "Location information" is information indicating the location where abnormal behavior was detected, and includes geographic coordinates, addresses, etc.

[0786] "API" stands for Application Program Interface, an interface that allows software functions and data to be used by other programs.

[0787] A "push notification system" is a system that sends notifications from a server to a user terminal in real time.

[0788] "Face mosaic processing" is a process of processing video data so that an individual's face cannot be identified, and includes polygon and blur effects.

[0789] "Audio masking" is a process of processing audio data to prevent individuals from being identified.

[0790] This invention describes a specific embodiment of a surveillance device video analysis system. This system is mainly composed of a server, a terminal, and a user, and aims to efficiently analyze video data acquired from surveillance devices and prevent crime.

[0791] First, the server acquires video data in real time from monitoring devices connected to the network (for example, fixed cameras or drive recorders). The server temporarily stores this video data in storage. To acquire the video data, for example, the video stream URL of an IP camera can be used.

[0792] The server then performs preprocessing on the acquired video data. This preprocessing includes noise reduction and frame rate normalization. For example, noise reduction using a Gaussian filter and standardizing the video frame rate to 30 fps are performed. This preprocessing allows for the efficient execution of subsequent analysis processes.

[0793] The server then sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the data and detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for abnormal behavior.

[0794] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, specifically by blurring faces (for example, using AWS Rekognition) and masking voices to remove personally identifiable information.

[0795] The server then notifies the user's device (e.g., smartphone or PC) and related systems of the analysis results. Notifications are sent using APIs or push notification systems. Users can receive notifications on their devices and take action. If necessary, the server also securely shares anonymized data with related organizations (e.g., police or security companies).

[0796] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device model based on new data to improve analysis accuracy. For example, TensorFlow or PyTorch can be used for this retraining.

[0797] Specific examples

[0798] Example of suspicious person detection in a commercial facility

[0799] Surveillance devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly. In this way, the surveillance device video analysis system efficiently detects abnormal behavior and acts quickly to deter crime. Privacy protection is also guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0800] Example prompts for generative AI models

[0801] text

[0802] Surveillance cameras installed in a commercial facility send video footage to a server in real time. The server preprocesses this video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the smartphone of the commercial facility manager. The manager receives the alert and responds promptly. Please explain this series of processes in easy-to-understand text.

[0803] This system makes it possible to efficiently analyze video data acquired from surveillance cameras, quickly detect and notify users of abnormal behavior, and provides a sense of security to users as it also takes privacy protection into consideration.

[0804] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0805] Step 1:

[0806] Acquiring and storing video data

[0807] The server acquires video data in real time from surveillance devices (e.g., fixed cameras or dashcams). This is done using the IP address or stream URL of the surveillance camera. The server then temporarily stores the acquired video data in storage. Specifically, it establishes a streaming connection for the video data and stores the acquired video frames sequentially.

[0808] Input: Surveillance video stream

[0809] Output: Temporarily saved video data

[0810] Step 2:

[0811] Video data preprocessing

[0812] The server performs preprocessing on the stored video data. This preprocessing includes noise reduction and frame rate normalization. A Gaussian filter is used for noise reduction, and frame rate normalization involves thinning and interpolating frames along the time axis. This improves data quality and increases the accuracy of analysis.

[0813] Input: Temporarily stored video data

[0814] Output: Pre-processed video data

[0815] Step 3:

[0816] Analysis of preprocessed video data

[0817] The server sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the video data and detect human movements and abnormal behavior. This analysis model operates in real time using pre-trained parameters.

[0818] Input: Preprocessed video data

[0819] Output: Analysis results (detection results of abnormal behavior, timestamp, location information)

[0820] Step 4:

[0821] Anonymization of analysis results

[0822] The server receives the analysis results and performs face blurring and audio masking to protect privacy. Face blurring is performed using image recognition software (e.g., AWS Rekognition), and audio masking is performed using technology that removes specific frequency bands.

[0823] Input: Analysis results

[0824] Output: Anonymized analysis results (face mosaic and voice masked data)

[0825] Step 5:

[0826] Notification to user devices and related systems

[0827] The server notifies the user's device and related systems of the anonymized analysis results using an API or push notification system. Users can receive the notifications on their smartphones or computers and respond promptly.

[0828] Input: Anonymized analysis results

[0829] Output: Notification to user device (push notification to smartphone or PC)

[0830] Step 6:

[0831] Saving analysis results and retraining models

[0832] The server stores the analysis results in a database and uses them as training data for the system. The analysis device model is periodically retrained based on new data to improve analysis accuracy. This retraining process uses TensorFlow and PyTorch to update the model parameters.

[0833] Input: Analysis results

[0834] Output: Updated analytical model, saved training data

[0835] These steps enable the system to achieve highly accurate detection of abnormal behavior, protect privacy, and support rapid response.

[0836] (Application example 1)

[0837] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0838] Conventional monitoring devices and systems often have delays in detecting abnormal behavior and subsequent responses, making it difficult to respond quickly. Another issue is that data privacy protection is insufficient, increasing the risk of personal information leaks. Furthermore, improving system training and analysis accuracy takes a lot of time and effort. There is a need for a system that can solve these issues and detect and respond to abnormal behavior more efficiently and quickly.

[0839] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0840] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect abnormal behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for notifying a smartphone application of abnormal behavior alerts in real time, and means for the user to view past abnormal behavior detection history. This enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases user peace of mind.

[0841] A "surveillance device" is a device for acquiring video data, such as a fixed camera or a drive recorder.

[0842] "Video data" refers to visual information acquired from a monitoring device, and is data used to detect abnormal behavior.

[0843] "Preprocessing" refers to the process of removing noise from the acquired video data, normalizing the frame rate, and otherwise processing the data to make it easier to analyze.

[0844] The "analysis device" is a device for analyzing pre-processed video data and detecting abnormal behavior.

[0845] "Abnormal behavior" refers to behavior that is different from the norm, such as the intrusion of suspicious individuals, violent behavior, or abnormal stopping of vehicles.

[0846] "Anonymization" is a process that removes information that can identify individuals from the analysis results in order to protect privacy.

[0847] A "user terminal" is a device, such as a smartphone or PC, that a user uses to check analysis results and receive notifications.

[0848] "Related systems" are external systems, such as police and security companies, that are necessary for sharing analysis results.

[0849] A "database" is a data storage device that stores analysis results and uses them as learning data for the system.

[0850] "Relearning" is the process of retraining the analysis device's model based on new data to improve analysis accuracy.

[0851] A "smartphone application" is software that runs on a smartphone and alerts users of abnormal behavior in real time.

[0852] An "abnormal behavior alert" is a warning or information that quickly notifies the user of detected abnormal behavior.

[0853] "History" is a record of abnormal behavior detected in the past, and is data including information that can be confirmed by the user.

[0854] A specific embodiment of a surveillance device video analysis system is described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from a surveillance device, detects abnormal behavior, and promptly notifies the user.

[0855] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored on storage (e.g., SSD or HDD). The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. For this purpose, libraries such as OpenCV are used. The processed video data is then sent to an analysis device.

[0856] The analysis device then receives the preprocessed video data and analyzes it using deep learning models (e.g., Keras or TensorFlow). This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0857] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. This includes blurring faces and masking audio to remove any personally identifiable information. The server then notifies the appropriate user device (e.g., smartphone, PC) and related systems (e.g., police, security companies) of the analysis results. Users receive alerts of abnormal behavior in real time through a smartphone application, allowing them to respond quickly. Users can also view past abnormal behavior detection history and consider countermeasures.

[0858] Furthermore, the server stores the analysis results in a database, which is used as training data for future systems. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. Retraining is performed using deep learning libraries such as Keras and TensorFlow.

[0859] As a specific example, consider a case where surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond quickly. In this way, the surveillance equipment video analysis system efficiently detects abnormal behavior and aims to deter crime by responding quickly.

[0860] Analysis using generative AI models uses prompts like the following:

[0861] "Analyze this video data to detect any abnormal behavior (e.g., unauthorized entry, violent behavior, or abnormal stopping). Include timestamps, detailed information, and location information in the detection results. Anonymize the data to protect privacy."

[0862] The above is a specific embodiment. This system enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases the sense of security of users.

[0863] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0864] Step 1:

[0865] The server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. The input is video data from the monitoring devices, and the output is data that is temporarily stored in storage. Specifically, the server receives video data sent from the monitoring devices and temporarily stores it.

[0866] Step 2:

[0867] The server performs preprocessing on the acquired video data, such as noise removal and frame rate normalization. The input here is the video data stored in storage, and the output is preprocessed, clean video data. Specifically, the server uses the OpenCV library to remove noise from the video and adjust the frame rate to a constant value.

[0868] Step 3:

[0869] The server sends the preprocessed video data to the analysis device. The input is the preprocessed video data, and the output is the data sent to the analysis device. Specifically, the preprocessed data is transferred to the analysis device via a network.

[0870] Step 4:

[0871] The analysis device receives the preprocessed video data and analyzes it using a deep learning model. This analysis detects abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The input is the preprocessed video data sent to the analysis device, and the output is the abnormal behavior detection results (timestamp, detailed information, and location information). Specifically, the data is analyzed using a generative AI model using Keras and TensorFlow.

[0872] Step 5:

[0873] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior detection results from the analysis device, and the output is anonymized data. Specific operations include face mosaic processing and voice masking to remove personally identifiable information.

[0874] Step 6:

[0875] The server notifies the user device (smartphone, PC) and related systems (police, security companies) of the anonymized analysis results. The input is the anonymized analysis results, and the output is the notified data. Specifically, the server sends the data to the user device or external system via an HTTP request, etc.

[0876] Step 7:

[0877] Users can receive alerts of abnormal behavior in real time through a smartphone application. The input is notification data from the server, and the output is alert information displayed on the smartphone application. Specifically, the application notification function receives the data and displays it on the screen.

[0878] Step 8:

[0879] Users can view the history of past abnormal behavior detections using a smartphone application. The input is the analysis results stored in the database, and the output is the history information displayed on the application. Specifically, the application queries the database, retrieves past detection results, and displays them.

[0880] Step 9:

[0881] The server stores the analysis results in a database and uses them as learning data for the system. The input is the analysis results, and the output is the data stored in the database. Specifically, the analysis results are formalized and stored in the database.

[0882] Step 10:

[0883] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is new data and past analysis results, and the output is the retrained model. Specifically, the model is updated using the retraining functions of Keras or TensorFlow.

[0884] This series of steps enables efficient and rapid detection and response to abnormal behavior, ensuring the safety of users.

[0885] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0886] Below, we will explain a specific embodiment that combines a surveillance video analysis system with an emotion engine. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to deter crime and recognize user emotions.

[0887] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[0888] The analysis device then receives the preprocessed video data and uses deep learning models to analyze human movements and abnormal behaviors (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behaviors.

[0889] Furthermore, the analysis device is equipped with an emotion engine that analyzes the user's facial expressions and tone of voice from the video data to identify their emotions. This emotion analysis can determine whether the user is expressing emotions such as fear, surprise, or anger.

[0890] Once the analysis is complete, the server retrieves the results and anonymizes them to protect privacy. It uses techniques like face blurring and voice masking to remove any personally identifiable information. Sentiment analysis results are also anonymized.

[0891] The server then notifies the appropriate user device and related systems of the analysis and sentiment analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take appropriate action. Additionally, the server can securely share anonymized data with relevant agencies (police, security companies) as needed.

[0892] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0893] Specific examples

[0894] Example of suspicious person detection in a commercial facility

[0895] Monitoring devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, the emotion engine analyzes the video data to determine the suspicious person's fear or tension, and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0896] In this way, a system that combines a surveillance video analysis system and an emotion engine can efficiently detect abnormal behavior and quickly respond to it, thereby preventing crime. Emotion analysis also enables early detection of potential danger, further enhancing safety. Privacy protection is also guaranteed, improving users' sense of security. This system can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[0897] The processing flow will be explained below.

[0898] Step 1:

[0899] The server receives video data in real time from network-connected surveillance devices, such as fixed cameras and dashcams, and connects to these devices to receive their signals.

[0900] Step 2:

[0901] The server temporarily stores the acquired video data in storage so that the data can be referenced later for processing.

[0902] Step 3:

[0903] The server performs preprocessing on the video data stored in the storage, including noise removal, frame rate normalization, and resolution adjustment.

[0904] Step 4:

[0905] The server transmits the pre-processed video data to the analysis equipment, and this transmission is fast and secure, ensuring that the data reaches the analysis equipment reliably.

[0906] Step 5:

[0907] The analysis device receives the preprocessed video data and uses deep learning models to analyze people and their behavior in the video, detecting abnormal behavior (such as intrusions, violent behavior, or abnormal vehicle stalls).

[0908] Step 6:

[0909] If the analysis device detects abnormal behavior, it records the timestamp and location information and sends this to the server as the analysis result.

[0910] Step 7:

[0911] The emotion engine included in the analysis device analyzes the user's facial expressions and tone of voice from the video data to generate emotion data, which can include fear, surprise, anger, etc.

[0912] Step 8:

[0913] The analysis device sends the emotion data to the server, where it is treated in the same way as the analysis results of abnormal behavior.

[0914] Step 9:

[0915] The server receives the analysis results and emotion data, and anonymizes them to protect privacy. It also applies mosaic processing to the video and masks the audio data.

[0916] Step 10:

[0917] The server notifies the user device and related systems of the anonymized analysis results and emotion data, and if notification is required, sends an alert to specific users or systems.

[0918] Step 11:

[0919] Users receive notifications on their smartphones or PCs and can take action as necessary. For example, a commercial facility manager can immediately contact security staff to check the situation and take appropriate measures.

[0920] Step 12:

[0921] The server stores the analysis results and emotion data in a database, which will be used as learning data for the system in the future.

[0922] Step 13:

[0923] The server periodically uses the saved data to retrain the analysis device's model, improving the model's analysis accuracy and enabling it to detect abnormal behavior and user emotions with greater accuracy.

[0924] Step 14:

[0925] The server applies the retrained model to the system and continues analyzing the video data from the surveillance device with improved analysis accuracy.

[0926] Example 2

[0927] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0928] Conventional surveillance systems were specialized in detecting abnormal behavior, but lacked the ability to analyze people's emotions, making it difficult to detect potential dangers early on. Furthermore, from the perspective of protecting privacy, it was necessary to adequately protect personally identifiable information.

[0929] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for acquiring video data from a monitoring device, a means for preprocessing the acquired video data, and a means for transmitting the preprocessed video data to an analysis device. This makes it possible to acquire video data in real time, detect abnormal behavior, and identify user emotions. The server also includes a means for anonymizing the analysis results and notifying the user terminal and related systems, and a means for storing the analysis results in a database and using them as learning data for the system. The server also includes a means for retraining the model of the analysis device based on new data to improve analysis accuracy. This makes it possible to continuously improve the analysis capabilities of the system while achieving both abnormal behavior detection accuracy and privacy protection.

[0930] A "surveillance device" is a device, such as a fixed camera or a drive recorder, that is installed to monitor a specific area.

[0931] "Video data" refers to visual information acquired from a surveillance device, typically consisting of a series of image frames.

[0932] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on the acquired video data to prepare it for analysis.

[0933] An "analysis device" is a device that uses advanced algorithms, such as deep learning models, to analyze pre-processed video data and detect abnormal behavior and emotions.

[0934] "Abnormal behavior" refers to movements or positions of people or objects that deviate from the normal movement of people or objects, and includes intrusions, violent acts, and abnormal stopping of vehicles.

[0935] The "emotion engine" is a system that analyzes a user's facial expressions and tone of voice from video data to identify emotions such as fear, surprise, and anger.

[0936] "Anonymization" refers to the process of removing personally identifiable information from analysis results to protect privacy. Specifically, it refers to processes such as blurring faces and masking audio.

[0937] A "user terminal" is a device, such as a smartphone or computer, that a user uses to receive information.

[0938] A "database" is a collection of data that stores analysis results and emotion analysis results and is used as learning data for the future.

[0939] "Model retraining" is the process of retraining the analysis device's algorithms based on new data to improve analysis accuracy.

[0940] The present invention relates to a system for acquiring video data from a monitoring device, analyzing the video data, and detecting abnormal behavior and emotions. Specific embodiments of the present invention will be described in detail below.

[0941] First, the server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. This video data is temporarily stored in storage on the server. The server then performs preprocessing on the acquired video data, performing noise removal and frame rate normalization. Software libraries such as OpenCV and FFmpeg are currently used for preprocessing, which are standard technologies.

[0942] The preprocessed video data is then sent to an analysis device, which uses machine learning libraries such as TensorFlow and PyTorch to analyze human movement and abnormal behavior using deep learning models. This analysis detects human movement and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[0943] The analysis device also incorporates an emotion engine, which analyzes the user's facial expressions and tone of voice from the video data to identify emotions. Using facial expression recognition algorithms and voice analysis techniques, it is possible to determine the emotions expressed by the user (e.g., fear, surprise, anger, etc.). It is recommended to use the Facial-Emotion-Recognition library for this emotion identification.

[0944] Once the analysis is complete, the server retrieves the results and anonymizes them, specifically by blurring faces and masking audio to remove any personally identifiable information. This can be done with the help of dlib or other facial recognition libraries.

[0945] The server then notifies the appropriate user device and related systems of the anonymized analysis results and sentiment analysis results. Users receive notifications on their smartphones, PCs, or other devices and can take action as needed. If necessary, the server also securely shares the anonymized data with relevant agencies (police, security companies). This is done via a secure API.

[0946] The server also stores the analysis results and sentiment analysis results in a database. This stored data is also used as a dataset for future learning. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. A batch training method is used for retraining, and the model is updated.

[0947] As a concrete example, consider a scenario in which surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, an emotion engine analyzes the suspicious person's fear and tension from the video data and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[0948] Through the above process, the system can efficiently detect abnormal behavior and respond quickly to deter crime, while also detecting potential dangers early through emotion analysis, further enhancing safety.

[0949] Example prompt sentence:

[0950] "Please explain the real-time suspicious person detection system for commercial facilities. Please include the names of the specific hardware and software, as well as the type of data processing that is performed."

[0951] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0952] Step 1:

[0953] The server receives video data in real time via the network from surveillance devices such as fixed cameras and dashcams. The input is the video data sent from the surveillance devices, and the output is the video data temporarily stored in the server's storage. This data is used in the next pre-processing step.

[0954] Step 2:

[0955] The server performs noise reduction and frame rate normalization on the acquired video data. The input is the video data saved in step 1, and the output is the preprocessed video data. The OpenCV library is used for noise reduction, and algorithms such as Gaussian Blur can be used. FFmpeg is used for frame rate normalization.

[0956] Step 3:

[0957] The server sends the preprocessed video data to the analysis device. The input is the video data preprocessed in step 2, and the output is the video data in the format required by the analysis device. This data transmission uses HTTP requests and socket communication.

[0958] Step 4:

[0959] The analysis device uses a deep learning model to analyze the preprocessed video data and detect abnormal behavior. The input is the video data sent in step 3, and the output is the analysis results of abnormal behavior (including timestamps and location information). Libraries such as TensorFlow and PyTorch are used for the analysis. Specifically, person detection algorithms such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector) are used.

[0960] Step 5:

[0961] The analysis device uses an emotion engine to analyze the user's emotions from the video data. The input is the same video data as in step 3, and the output is the emotion analysis results (emotional information such as fear, surprise, and anger). The Facial-Emotion-Recognition library is used for emotion analysis. Specific operations include a facial landmark detection algorithm and a facial expression classification algorithm.

[0962] Step 6:

[0963] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior analysis results and emotion analysis results obtained in steps 4 and 5, and the output is anonymized data. Specifically, the dlib library is used for face mosaic processing, and audio editing software such as Audacity is used for audio masking processing.

[0964] Step 7:

[0965] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized data from step 6, and the output is a notification message displayed on the user's device. This notification is sent using a push notification service such as Firebase Cloud Messaging or Apple Push Notification Service. Specifically, the server creates a notification message and sends it to the user's smartphone or PC.

[0966] Step 8:

[0967] The server stores the analysis results and sentiment analysis results in a database. The input is the analysis results and sentiment analysis results obtained in steps 4 and 5, and the output is the data stored in the database. This data will also be used as a training dataset in the future. Specifically, the server stores the data using an SQL database or NoSQL database.

[0968] Step 9:

[0969] The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. The input is the training data stored in the database, and the output is the retrained analysis model. Specific operations include batch training and updating the model weights. Possible libraries used are TensorFlow and PyTorch.

[0970] (Application example 2)

[0971] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0972] In modern society, security threats are increasing daily, and there is a need for rapid and accurate detection and response of abnormal behavior, especially in large commercial and public facilities. While numerous surveillance cameras are installed in such locations, it is difficult to monitor all camera footage in real time due to limited human resources. Furthermore, emotion analysis is also important in detecting abnormal behavior, making it necessary to discover and respond to potential threats early. Furthermore, there is a need for a method to solve these issues while protecting individual privacy.

[0973] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0974] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for the analysis device to analyze a person's emotions, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying a user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, and means for displaying the notified analysis results on a user terminal. This enables security staff to check anomalous behavior and emotion analysis results in real time and quickly recognize and respond to potential threats.

[0975] "Monitoring equipment" refers to equipment used to acquire video data, and primarily refers to fixed cameras and dashcams.

[0976] "Video data" refers to visual information acquired from a monitoring device, and is digital data that records movements and situations.

[0977] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on video data.

[0978] An "analysis device" is a device that uses preprocessed video data to analyze people's movements, abnormal behavior, and emotions.

[0979] "Abnormal behavior" refers to any unusual behavior or action, including, for example, intrusion, violent behavior, or abnormal stopping of a vehicle.

[0980] "Emotion analysis" is a technology that analyzes a person's facial expressions and tone of voice to identify specific emotions.

[0981] "Anonymization" is a process that removes personally identifiable information from the analysis results to protect the privacy of the data.

[0982] "User terminal" refers to electronic devices used to display analysis results, such as smartphones, PCs, and smart glasses.

[0983] "Related systems" refers to other systems or institutions that receive the analysis results, including police and security companies.

[0984] A "database" is a system for systematically storing and managing analysis results and learning data.

[0985] "Relearning" is a process of updating the model of the analysis device based on new data to improve the accuracy of the analysis.

[0986] "Notification" refers to the act of notifying the analysis results to the appropriate user terminal or related system.

[0987] The system according to the present invention can be implemented as a smart glasses application related to security services, and includes a monitoring device, a server, an analysis device, smart glasses, and related systems.

[0988] First, the server acquires video data in real time from monitoring devices (such as fixed cameras) connected to the network. This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to the analysis device.

[0989] The analyzer then receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). Furthermore, the analyzer incorporates an emotion engine that analyzes human facial expressions and speech tones in the video data to identify emotions. This emotion analysis can determine whether a person is displaying emotions such as fear, surprise, or anger.

[0990] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. Personally identifiable information is removed by methods such as facial mosaic and audio masking. Emotion analysis results are also anonymized. The server then notifies the appropriate user device and related systems of the analysis and emotion analysis results. Users receive notifications on their smart glasses and can check the results in real time. If necessary, the server also securely shares the anonymized data with relevant authorities (police, security companies).

[0991] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[0992] As a concrete example, if security staff at a large shopping mall are patrolling wearing smart glasses, images from surveillance cameras will be displayed on the glasses in real time. At the same time, the system can detect intrusions and any abnormal behavior, and the emotions of individuals who exhibit suspicious behavior (such as fear or tension) can also be displayed on the glasses. This allows security staff to quickly recognize potential threats and respond efficiently.

[0993] Here is an example of a prompt to input to a generative AI model:

[0994] Run the following code to create an application that displays abnormal behavior and emotion analysis results in real time on smart glasses worn by security staff at commercial facilities.

[0995] The surveillance equipment video analysis system detects abnormal behavior and emotions in real time. The detected information is displayed on the security staff's smart glasses. Specifically, the system performs noise removal and frame rate normalization on the video data, abnormal behavior analysis using a deep learning model, and emotion analysis using an emotion engine, and displays the results on the smart glasses.

[0996] Libraries used:

[0997] cv2 (OpenCV)

[0998] Deep learning models (e.g., DeepFace)

[0999] Smart Glasses API Library

[1000] Required functions:

[1001] get_camera_feed(): Gets video data from the server

[1002] preprocess_frame(frame): Preprocesses video data

[1003] detect_anomalies(frame): Detect abnormal behavior

[1004] analyze_emotions(frame): Analyze emotions

[1005] display_on_glasses(info): Display information on smart glasses

[1006] Please generate Python code that meets the above specifications.

[1007] As such, this system has a wide range of applications in the security field, particularly in terms of its ability to detect abnormal behavior in real time and respond quickly through emotion analysis.

[1008] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1009] Step 1:

[1010] The server acquires video data in real time from a monitoring device (such as a fixed camera). In this step, the server receives the video data sent from the monitoring device as streaming data and temporarily stores it in storage.

[1011] Step 2:

[1012] The server performs noise reduction and frame rate normalization on the acquired video data. The input is raw video data stored in storage, and the server applies noise reduction filters and frame rate unification processing to it before outputting the preprocessed video data.

[1013] Step 3:

[1014] The server transmits the preprocessed video data to the analysis device. The input is the preprocessed video data, and the server transmits this to the analysis device via the network.

[1015] Step 4:

[1016] The analysis device receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. The input of this step is the preprocessed video data, and the output is the detected abnormal behavior (e.g., intrusion, violent behavior, abnormal car stalling, etc.) along with its timestamp and location information.

[1017] Step 5:

[1018] The analyzer identifies emotions from video data by analyzing facial expressions and tone of speech. The input is pre-processed video data, and the output is the identified emotion (e.g., fear, surprise, anger, etc.). The analyzer does this using an emotion engine.

[1019] Step 6:

[1020] The server receives the analysis results and emotion analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior and emotion analysis results provided by the analysis device, and the output is the anonymized analysis results. The server performs operations such as face mosaic processing and audio masking.

[1021] Step 7:

[1022] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized analysis results, which the server notifies at the appropriate time. This allows the results to be displayed in real time on the user device (such as smart glasses).

[1023] Step 8:

[1024] The server stores the analysis results and sentiment analysis results in a database. The inputs to this step are the analysis results and sentiment analysis results, and the output is the stored data. The stored data can also be used as learning data in the future.

[1025] Step 9:

[1026] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is the newly saved data, which the server uses to retrain the deep learning model. The output is an analysis model with improved accuracy.

[1027] The detailed inputs, processing, and outputs for each step are shown below:

[1028] Step 1:

[1029] Input: Real-time video data from a surveillance device

[1030] Processing: The server receives the video data and temporarily stores it in storage.

[1031] Output: Raw video data stored in storage

[1032] Step 2:

[1033] Input: Raw video data stored in storage

[1034] Processing: Noise reduction filter applied, frame rate unified processing

[1035] Output: Pre-processed video data

[1036] Step 3:

[1037] Input: Preprocessed video data

[1038] Processing: Send data to an analysis device via a network

[1039] Output: Video data transmission to analysis device completed

[1040] Step 4:

[1041] Input: Preprocessed video data

[1042] Processing: Analyzing movement and abnormal behavior with deep learning models

[1043] Output: Detected abnormal behavior (timestamp and location information)

[1044] Step 5:

[1045] Input: Preprocessed video data

[1046] Processing: Identifying emotions with the emotion engine

[1047] Output: Identified emotion (e.g., fear, surprise, anger)

[1048] Step 6:

[1049] Input: Abnormal behavior and emotion analysis results

[1050] Processing: Anonymization by face mosaic processing, voice masking, etc.

[1051] Output: Anonymized analysis results

[1052] Step 7:

[1053] Input: Anonymized analysis results

[1054] Processing: Notify user terminals and related systems

[1055] Output: Notification to user terminal and related systems completed

[1056] Step 8:

[1057] Input: Analysis results and sentiment analysis results

[1058] Process: Save to database

[1059] Output: Data stored in the database

[1060] Step 9:

[1061] Input: Newly saved data

[1062] Processing: Retraining a deep learning model

[1063] Output: Improved analytical model

[1064] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1065] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1066] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1067] [Fourth embodiment]

[1068] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1069] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1070] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1071] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1072] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1073] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1074] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1075] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1076] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1077] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1078] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1079] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1080] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1081] A specific embodiment of a surveillance device video analysis system will be described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to prevent crime.

[1082] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[1083] The analysis device then receives the preprocessed video data and analyzes it using deep learning models to detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[1084] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, removing any personally identifiable information by pixelating faces and masking voices.

[1085] The server then notifies the appropriate user device and related systems of the analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take action. Additionally, the server can securely share anonymized data with relevant authorities (police, security companies) as needed.

[1086] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[1087] Specific examples

[1088] Example of suspicious person detection in a commercial facility

[1089] Monitoring equipment installed in commercial facilities transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly.

[1090] In this way, surveillance video analysis systems can efficiently detect abnormal behavior and quickly respond to crime, thereby preventing crime. Furthermore, privacy protection is guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[1091] The processing flow will be explained below.

[1092] Step 1:

[1093] The server acquires video data in real time from monitoring devices connected to the network. The monitoring devices come in a variety of forms, such as fixed cameras and dashcams, and the server connects to these devices to collect data.

[1094] Step 2:

[1095] The server temporarily stores the acquired video data in storage. This storage process is performed to back up the data so that it is not lost.

[1096] Step 3:

[1097] The server then begins pre-processing the video data stored in the storage, which includes noise removal, frame rate normalization, and resolution adjustment.

[1098] Step 4:

[1099] The server transmits the pre-processed video data to the analysis device, which transmits the data quickly and securely and ensures that the analysis device receives the data.

[1100] Step 5:

[1101] The analyzer receives the preprocessed data and uses deep learning models to analyze people's movements and abnormal behavior. In this process, deep learning algorithms detect abnormal behavior (such as intrusions or violent behavior) in the video.

[1102] Step 6:

[1103] When the analysis device detects abnormal behavior, it records the timestamp and location information and sends the analysis results to the server.

[1104] Step 7:

[1105] The server receives the analysis results, anonymizes any information that could identify individuals to protect privacy, and applies mosaic processing and audio masking to the video data.

[1106] Step 8:

[1107] The server notifies the user and related systems of the anonymized analysis results, and the user receives the notification on their smartphone or PC and can take action.

[1108] Step 9:

[1109] Users can check the notification and take immediate action if necessary. For example, a commercial facility manager can work with security staff to investigate the situation and take appropriate action.

[1110] Step 10:

[1111] The server stores the analysis results in a database, which will be used as learning data in the future.

[1112] Step 11:

[1113] The server periodically uses the stored data to retrain the analysis device's model, improving analysis accuracy and enabling more effective detection of abnormal behavior.

[1114] Step 12:

[1115] The server applies the retrained model to the system and continues analyzing the monitoring device data with the new analytical accuracy.

[1116] Example 1

[1117] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1118] Conventional surveillance systems have a problem in that they have low accuracy in analyzing video data, making it difficult to accurately detect abnormal behavior. Furthermore, they lack the functionality to detect and notify abnormal behavior in real time while protecting individual privacy, which often results in delayed response. To solve these problems, an efficient and highly accurate video data analysis system is required.

[1119] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1120] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for performing noise reduction and frame rate normalization on the preprocessed video data, means for recording timestamps and location information of anomalous behavior detected by the analysis device, means for performing face mosaic processing and audio masking when acquiring the analysis results and anonymizing the data for privacy protection, means for notifying the user terminal of the analysis results using an API or push notification system, and means for acquiring video data from the monitoring device in real time, preprocessing the video data, and transferring the video data to the analysis device, thereby enabling highly accurate detection of anomalous behavior.

[1121] A "monitoring device" is a device that continuously captures images of a monitored area and generates video data, and specifically includes fixed cameras and drive recorders.

[1122] "Video data" refers to video information acquired by a monitoring device and is made up of successive frames of images.

[1123] "Preprocessing" refers to processing such as noise removal and frame rate normalization before analyzing video data.

[1124] The "analysis device" is a device that runs a deep learning model to detect abnormal behavior using preprocessed video data as input.

[1125] "Abnormal behavior" refers to any unusual behavior or condition that occurs within the monitored area, including, for example, intrusion, violent behavior, or abnormal vehicle stalls.

[1126] "Anonymization" is a process carried out to protect information that could identify an individual from the analysis results, and specifically includes blurring faces and masking audio.

[1127] A "user terminal" is an electronic device capable of receiving and displaying analysis results, such as a smartphone or a personal computer.

[1128] A "database" is a system for efficiently storing and managing analysis results and learning data.

[1129] "Retraining" is the process of updating a deep learning model based on new data to improve analysis accuracy.

[1130] "Noise reduction" is the process of removing unnecessary background information and noise from video data.

[1131] "Frame rate normalization" is a process of standardizing the frame rate of video data to a fixed value.

[1132] A "timestamp" is information indicating a specific time, and is recorded in video data or analysis results.

[1133] "Location information" is information indicating the location where abnormal behavior was detected, and includes geographic coordinates, addresses, etc.

[1134] "API" stands for Application Program Interface, an interface that allows software functions and data to be used by other programs.

[1135] A "push notification system" is a system that sends notifications from a server to a user terminal in real time.

[1136] "Face mosaic processing" is a process of processing video data so that an individual's face cannot be identified, and includes polygon and blur effects.

[1137] "Audio masking" is a process of processing audio data to prevent individuals from being identified.

[1138] This invention describes a specific embodiment of a surveillance device video analysis system. This system is mainly composed of a server, a terminal, and a user, and aims to efficiently analyze video data acquired from surveillance devices and prevent crime.

[1139] First, the server acquires video data in real time from monitoring devices connected to the network (for example, fixed cameras or drive recorders). The server temporarily stores this video data in storage. To acquire the video data, for example, the video stream URL of an IP camera can be used.

[1140] The server then performs preprocessing on the acquired video data. This preprocessing includes noise reduction and frame rate normalization. For example, noise reduction using a Gaussian filter and standardizing the video frame rate to 30 fps are performed. This preprocessing allows for the efficient execution of subsequent analysis processes.

[1141] The server then sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the data and detect human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for abnormal behavior.

[1142] Once the analysis is complete, the server retrieves the results and anonymizes the data to protect privacy, specifically by blurring faces (for example, using AWS Rekognition) and masking voices to remove personally identifiable information.

[1143] The server then notifies the user's device (e.g., smartphone or PC) and related systems of the analysis results. Notifications are sent using APIs or push notification systems. Users can receive notifications on their devices and take action. If necessary, the server also securely shares anonymized data with related organizations (e.g., police or security companies).

[1144] Furthermore, the server stores the analysis results in a database. This stored data is used as future training data for the system. The server periodically retrains the analysis device model based on new data to improve analysis accuracy. For example, TensorFlow or PyTorch can be used for this retraining.

[1145] Specific examples

[1146] Example of suspicious person detection in a commercial facility

[1147] Surveillance devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and responds promptly. In this way, the surveillance device video analysis system efficiently detects abnormal behavior and acts quickly to deter crime. Privacy protection is also guaranteed, increasing users' sense of security. Such systems can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[1148] Example prompts for generative AI models

[1149] text

[1150] Surveillance cameras installed in a commercial facility send video footage to a server in real time. The server preprocesses this video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious person and sends the information to the server. The server anonymizes the analysis results and sends an alert notification to the smartphone of the commercial facility manager. The manager receives the alert and responds promptly. Please explain this series of processes in easy-to-understand text.

[1151] This system makes it possible to efficiently analyze video data acquired from surveillance cameras, quickly detect and notify users of abnormal behavior, and provides a sense of security to users as it also takes privacy protection into consideration.

[1152] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1153] Step 1:

[1154] Acquiring and storing video data

[1155] The server acquires video data in real time from surveillance devices (e.g., fixed cameras or dashcams). This is done using the IP address or stream URL of the surveillance camera. The server then temporarily stores the acquired video data in storage. Specifically, it establishes a streaming connection for the video data and stores the acquired video frames sequentially.

[1156] Input: Surveillance video stream

[1157] Output: Temporarily saved video data

[1158] Step 2:

[1159] Video data preprocessing

[1160] The server performs preprocessing on the stored video data. This preprocessing includes noise reduction and frame rate normalization. A Gaussian filter is used for noise reduction, and frame rate normalization involves thinning and interpolating frames along the time axis. This improves data quality and increases the accuracy of analysis.

[1161] Input: Temporarily stored video data

[1162] Output: Pre-processed video data

[1163] Step 3:

[1164] Analysis of preprocessed video data

[1165] The server sends the preprocessed video data to an analysis device, which uses a deep learning model (e.g., the YOLO model) to analyze the video data and detect human movements and abnormal behavior. This analysis model operates in real time using pre-trained parameters.

[1166] Input: Preprocessed video data

[1167] Output: Analysis results (detection results of abnormal behavior, timestamp, location information)

[1168] Step 4:

[1169] Anonymization of analysis results

[1170] The server receives the analysis results and performs face blurring and audio masking to protect privacy. Face blurring is performed using image recognition software (e.g., AWS Rekognition), and audio masking is performed using technology that removes specific frequency bands.

[1171] Input: Analysis results

[1172] Output: Anonymized analysis results (face mosaic and voice masked data)

[1173] Step 5:

[1174] Notification to user devices and related systems

[1175] The server notifies the user's device and related systems of the anonymized analysis results using an API or push notification system. Users can receive the notifications on their smartphones or computers and respond promptly.

[1176] Input: Anonymized analysis results

[1177] Output: Notification to user device (push notification to smartphone or PC)

[1178] Step 6:

[1179] Saving analysis results and retraining models

[1180] The server stores the analysis results in a database and uses them as training data for the system. The analysis device model is periodically retrained based on new data to improve analysis accuracy. This retraining process uses TensorFlow and PyTorch to update the model parameters.

[1181] Input: Analysis results

[1182] Output: Updated analytical model, saved training data

[1183] These steps enable the system to achieve highly accurate detection of abnormal behavior, protect privacy, and support rapid response.

[1184] (Application example 1)

[1185] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1186] Conventional monitoring devices and systems often have delays in detecting abnormal behavior and subsequent responses, making it difficult to respond quickly. Another issue is that data privacy protection is insufficient, increasing the risk of personal information leaks. Furthermore, improving system training and analysis accuracy takes a lot of time and effort. There is a need for a system that can solve these issues and detect and respond to abnormal behavior more efficiently and quickly.

[1187] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1188] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect abnormal behavior, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying the user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, means for notifying a smartphone application of abnormal behavior alerts in real time, and means for the user to view past abnormal behavior detection history. This enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases user peace of mind.

[1189] A "surveillance device" is a device for acquiring video data, such as a fixed camera or a drive recorder.

[1190] "Video data" refers to visual information acquired from a monitoring device, and is data used to detect abnormal behavior.

[1191] "Preprocessing" refers to the process of removing noise from the acquired video data, normalizing the frame rate, and otherwise processing the data to make it easier to analyze.

[1192] The "analysis device" is a device for analyzing pre-processed video data and detecting abnormal behavior.

[1193] "Abnormal behavior" refers to behavior that is different from the norm, such as the intrusion of suspicious individuals, violent behavior, or abnormal stopping of vehicles.

[1194] "Anonymization" is a process that removes information that can identify individuals from the analysis results in order to protect privacy.

[1195] A "user terminal" is a device, such as a smartphone or PC, that a user uses to check analysis results and receive notifications.

[1196] "Related systems" are external systems, such as police and security companies, that are necessary for sharing analysis results.

[1197] A "database" is a data storage device that stores analysis results and uses them as learning data for the system.

[1198] "Relearning" is the process of retraining the analysis device's model based on new data to improve analysis accuracy.

[1199] A "smartphone application" is software that runs on a smartphone and alerts users of abnormal behavior in real time.

[1200] An "abnormal behavior alert" is a warning or information that quickly notifies the user of detected abnormal behavior.

[1201] "History" is a record of abnormal behavior detected in the past, and is data including information that can be confirmed by the user.

[1202] A specific embodiment of a surveillance device video analysis system is described below. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from a surveillance device, detects abnormal behavior, and promptly notifies the user.

[1203] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored on storage (e.g., SSD or HDD). The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. For this purpose, libraries such as OpenCV are used. The processed video data is then sent to an analysis device.

[1204] The analysis device then receives the preprocessed video data and analyzes it using deep learning models (e.g., Keras or TensorFlow). This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[1205] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. This includes blurring faces and masking audio to remove any personally identifiable information. The server then notifies the appropriate user device (e.g., smartphone, PC) and related systems (e.g., police, security companies) of the analysis results. Users receive alerts of abnormal behavior in real time through a smartphone application, allowing them to respond quickly. Users can also view past abnormal behavior detection history and consider countermeasures.

[1206] Furthermore, the server stores the analysis results in a database, which is used as training data for future systems. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. Retraining is performed using deep learning libraries such as Keras and TensorFlow.

[1207] As a specific example, consider a case where surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects the intrusion of a suspicious individual and sends the information to the server. The server anonymizes the analysis results and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond quickly. In this way, the surveillance equipment video analysis system efficiently detects abnormal behavior and aims to deter crime by responding quickly.

[1208] Analysis using generative AI models uses prompts like the following:

[1209] "Analyze this video data to detect any abnormal behavior (e.g., unauthorized entry, violent behavior, or abnormal stopping). Include timestamps, detailed information, and location information in the detection results. Anonymize the data to protect privacy."

[1210] The above is a specific embodiment. This system enables efficient and rapid detection and response to abnormal behavior, ensures privacy protection, and increases the sense of security of users.

[1211] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1212] Step 1:

[1213] The server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. The input is video data from the monitoring devices, and the output is data that is temporarily stored in storage. Specifically, the server receives video data sent from the monitoring devices and temporarily stores it.

[1214] Step 2:

[1215] The server performs preprocessing on the acquired video data, such as noise removal and frame rate normalization. The input here is the video data stored in storage, and the output is preprocessed, clean video data. Specifically, the server uses the OpenCV library to remove noise from the video and adjust the frame rate to a constant value.

[1216] Step 3:

[1217] The server sends the preprocessed video data to the analysis device. The input is the preprocessed video data, and the output is the data sent to the analysis device. Specifically, the preprocessed data is transferred to the analysis device via a network.

[1218] Step 4:

[1219] The analysis device receives the preprocessed video data and analyzes it using a deep learning model. This analysis detects abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The input is the preprocessed video data sent to the analysis device, and the output is the abnormal behavior detection results (timestamp, detailed information, and location information). Specifically, the data is analyzed using a generative AI model using Keras and TensorFlow.

[1220] Step 5:

[1221] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior detection results from the analysis device, and the output is anonymized data. Specific operations include face mosaic processing and voice masking to remove personally identifiable information.

[1222] Step 6:

[1223] The server notifies the user device (smartphone, PC) and related systems (police, security companies) of the anonymized analysis results. The input is the anonymized analysis results, and the output is the notified data. Specifically, the server sends the data to the user device or external system via an HTTP request, etc.

[1224] Step 7:

[1225] Users can receive alerts of abnormal behavior in real time through a smartphone application. The input is notification data from the server, and the output is alert information displayed on the smartphone application. Specifically, the application notification function receives the data and displays it on the screen.

[1226] Step 8:

[1227] Users can view the history of past abnormal behavior detections using a smartphone application. The input is the analysis results stored in the database, and the output is the history information displayed on the application. Specifically, the application queries the database, retrieves past detection results, and displays them.

[1228] Step 9:

[1229] The server stores the analysis results in a database and uses them as learning data for the system. The input is the analysis results, and the output is the data stored in the database. Specifically, the analysis results are formalized and stored in the database.

[1230] Step 10:

[1231] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is new data and past analysis results, and the output is the retrained model. Specifically, the model is updated using the retraining functions of Keras or TensorFlow.

[1232] This series of steps enables efficient and rapid detection and response to abnormal behavior, ensuring the safety of users.

[1233] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1234] Below, we will explain a specific embodiment that combines a surveillance video analysis system with an emotion engine. This system is mainly composed of a server, a terminal, and a user, and efficiently analyzes video data acquired from surveillance devices to deter crime and recognize user emotions.

[1235] First, the server acquires video data in real time from network-connected surveillance devices (such as fixed cameras and dashcams). This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to an analysis device.

[1236] The analysis device then receives the preprocessed video data and uses deep learning models to analyze human movements and abnormal behaviors (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behaviors.

[1237] Furthermore, the analysis device is equipped with an emotion engine that analyzes the user's facial expressions and tone of voice from the video data to identify their emotions. This emotion analysis can determine whether the user is expressing emotions such as fear, surprise, or anger.

[1238] Once the analysis is complete, the server retrieves the results and anonymizes them to protect privacy. It uses techniques like face blurring and voice masking to remove any personally identifiable information. Sentiment analysis results are also anonymized.

[1239] The server then notifies the appropriate user device and related systems of the analysis and sentiment analysis results. Users can receive notifications on their smartphones, PCs, or other devices and take appropriate action. Additionally, the server can securely share anonymized data with relevant agencies (police, security companies) as needed.

[1240] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[1241] Specific examples

[1242] Example of suspicious person detection in a commercial facility

[1243] Monitoring devices installed in commercial facilities send video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, the emotion engine analyzes the video data to determine the suspicious person's fear or tension, and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[1244] In this way, a system that combines a surveillance video analysis system and an emotion engine can efficiently detect abnormal behavior and quickly respond to it, thereby preventing crime. Emotion analysis also enables early detection of potential danger, further enhancing safety. Privacy protection is also guaranteed, improving users' sense of security. This system can be implemented in a wide range of locations, including commercial facilities, public facilities, transportation facilities, and residential areas.

[1245] The processing flow will be explained below.

[1246] Step 1:

[1247] The server receives video data in real time from network-connected surveillance devices, such as fixed cameras and dashcams, and connects to these devices to receive their signals.

[1248] Step 2:

[1249] The server temporarily stores the acquired video data in storage so that the data can be referenced later for processing.

[1250] Step 3:

[1251] The server performs preprocessing on the video data stored in the storage, including noise removal, frame rate normalization, and resolution adjustment.

[1252] Step 4:

[1253] The server transmits the pre-processed video data to the analysis equipment, and this transmission is fast and secure, ensuring that the data reaches the analysis equipment reliably.

[1254] Step 5:

[1255] The analysis device receives the preprocessed video data and uses deep learning models to analyze people and their behavior in the video, detecting abnormal behavior (such as intrusions, violent behavior, or abnormal vehicle stalls).

[1256] Step 6:

[1257] If the analysis device detects abnormal behavior, it records the timestamp and location information and sends this to the server as the analysis result.

[1258] Step 7:

[1259] The emotion engine included in the analysis device analyzes the user's facial expressions and tone of voice from the video data to generate emotion data, which can include fear, surprise, anger, etc.

[1260] Step 8:

[1261] The analysis device sends the emotion data to the server, where it is treated in the same way as the analysis results of abnormal behavior.

[1262] Step 9:

[1263] The server receives the analysis results and emotion data, and anonymizes them to protect privacy. It also applies mosaic processing to the video and masks the audio data.

[1264] Step 10:

[1265] The server notifies the user device and related systems of the anonymized analysis results and emotion data, and if notification is required, sends an alert to specific users or systems.

[1266] Step 11:

[1267] Users receive notifications on their smartphones or PCs and can take action as necessary. For example, a commercial facility manager can immediately contact security staff to check the situation and take appropriate measures.

[1268] Step 12:

[1269] The server stores the analysis results and emotion data in a database, which will be used as learning data for the system in the future.

[1270] Step 13:

[1271] The server periodically uses the saved data to retrain the analysis device's model, improving the model's analysis accuracy and enabling it to detect abnormal behavior and user emotions with greater accuracy.

[1272] Step 14:

[1273] The server applies the retrained model to the system and continues analyzing the video data from the surveillance device with improved analysis accuracy.

[1274] Example 2

[1275] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1276] Conventional surveillance systems were specialized in detecting abnormal behavior, but lacked the ability to analyze people's emotions, making it difficult to detect potential dangers early on. Furthermore, from the perspective of protecting privacy, it was necessary to adequately protect personally identifiable information.

[1277] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes a means for acquiring video data from a monitoring device, a means for preprocessing the acquired video data, and a means for transmitting the preprocessed video data to an analysis device. This makes it possible to acquire video data in real time, detect abnormal behavior, and identify user emotions. The server also includes a means for anonymizing the analysis results and notifying the user terminal and related systems, and a means for storing the analysis results in a database and using them as learning data for the system. The server also includes a means for retraining the model of the analysis device based on new data to improve analysis accuracy. This makes it possible to continuously improve the analysis capabilities of the system while achieving both abnormal behavior detection accuracy and privacy protection.

[1278] A "surveillance device" is a device, such as a fixed camera or a drive recorder, that is installed to monitor a specific area.

[1279] "Video data" refers to visual information acquired from a surveillance device, typically consisting of a series of image frames.

[1280] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on the acquired video data to prepare it for analysis.

[1281] An "analysis device" is a device that uses advanced algorithms, such as deep learning models, to analyze pre-processed video data and detect abnormal behavior and emotions.

[1282] "Abnormal behavior" refers to movements or positions of people or objects that deviate from the normal movement of people or objects, and includes intrusions, violent acts, and abnormal stopping of vehicles.

[1283] The "emotion engine" is a system that analyzes a user's facial expressions and tone of voice from video data to identify emotions such as fear, surprise, and anger.

[1284] "Anonymization" refers to the process of removing personally identifiable information from analysis results to protect privacy. Specifically, it refers to processes such as blurring faces and masking audio.

[1285] A "user terminal" is a device, such as a smartphone or computer, that a user uses to receive information.

[1286] A "database" is a collection of data that stores analysis results and emotion analysis results and is used as learning data for the future.

[1287] "Model retraining" is the process of retraining the analysis device's algorithms based on new data to improve analysis accuracy.

[1288] The present invention relates to a system for acquiring video data from a monitoring device, analyzing the video data, and detecting abnormal behavior and emotions. Specific embodiments of the present invention will be described in detail below.

[1289] First, the server acquires video data in real time from monitoring devices (such as fixed cameras and dashcams) connected to the network. This video data is temporarily stored in storage on the server. The server then performs preprocessing on the acquired video data, performing noise removal and frame rate normalization. Software libraries such as OpenCV and FFmpeg are currently used for preprocessing, which are standard technologies.

[1290] The preprocessed video data is then sent to an analysis device, which uses machine learning libraries such as TensorFlow and PyTorch to analyze human movement and abnormal behavior using deep learning models. This analysis detects human movement and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). The analysis results include timestamps and location information for detected abnormal behavior.

[1291] The analysis device also incorporates an emotion engine, which analyzes the user's facial expressions and tone of voice from the video data to identify emotions. Using facial expression recognition algorithms and voice analysis techniques, it is possible to determine the emotions expressed by the user (e.g., fear, surprise, anger, etc.). It is recommended to use the Facial-Emotion-Recognition library for this emotion identification.

[1292] Once the analysis is complete, the server retrieves the results and anonymizes them, specifically by blurring faces and masking audio to remove any personally identifiable information. This can be done with the help of dlib or other facial recognition libraries.

[1293] The server then notifies the appropriate user device and related systems of the anonymized analysis results and sentiment analysis results. Users receive notifications on their smartphones, PCs, or other devices and can take action as needed. If necessary, the server also securely shares the anonymized data with relevant agencies (police, security companies). This is done via a secure API.

[1294] The server also stores the analysis results and sentiment analysis results in a database. This stored data is also used as a dataset for future learning. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. A batch training method is used for retraining, and the model is updated.

[1295] As a concrete example, consider a scenario in which surveillance equipment installed in a commercial facility transmits video footage to a server in real time. The server preprocesses the video data and forwards it to an analysis device. The analysis device detects suspicious intrusions and abnormal behavior and sends this information to the server. At the same time, an emotion engine analyzes the suspicious person's fear and tension from the video data and sends this information to the server. The server anonymizes the analysis results and emotion data and sends an alert notification to the commercial facility manager and security office. The manager receives the alert on their smartphone and can respond promptly.

[1296] Through the above process, the system can efficiently detect abnormal behavior and respond quickly to deter crime, while also detecting potential dangers early through emotion analysis, further enhancing safety.

[1297] Example prompt sentence:

[1298] "Please explain the real-time suspicious person detection system for commercial facilities. Please include the names of the specific hardware and software, as well as the type of data processing that is performed."

[1299] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1300] Step 1:

[1301] The server receives video data in real time via the network from surveillance devices such as fixed cameras and dashcams. The input is the video data sent from the surveillance devices, and the output is the video data temporarily stored in the server's storage. This data is used in the next pre-processing step.

[1302] Step 2:

[1303] The server performs noise reduction and frame rate normalization on the acquired video data. The input is the video data saved in step 1, and the output is the preprocessed video data. The OpenCV library is used for noise reduction, and algorithms such as Gaussian Blur can be used. FFmpeg is used for frame rate normalization.

[1304] Step 3:

[1305] The server sends the preprocessed video data to the analysis device. The input is the video data preprocessed in step 2, and the output is the video data in the format required by the analysis device. This data transmission uses HTTP requests and socket communication.

[1306] Step 4:

[1307] The analysis device uses a deep learning model to analyze the preprocessed video data and detect abnormal behavior. The input is the video data sent in step 3, and the output is the analysis results of abnormal behavior (including timestamps and location information). Libraries such as TensorFlow and PyTorch are used for the analysis. Specifically, person detection algorithms such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector) are used.

[1308] Step 5:

[1309] The analysis device uses an emotion engine to analyze the user's emotions from the video data. The input is the same video data as in step 3, and the output is the emotion analysis results (emotional information such as fear, surprise, and anger). The Facial-Emotion-Recognition library is used for emotion analysis. Specific operations include a facial landmark detection algorithm and a facial expression classification algorithm.

[1310] Step 6:

[1311] The server receives the analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior analysis results and emotion analysis results obtained in steps 4 and 5, and the output is anonymized data. Specifically, the dlib library is used for face mosaic processing, and audio editing software such as Audacity is used for audio masking processing.

[1312] Step 7:

[1313] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized data from step 6, and the output is a notification message displayed on the user's device. This notification is sent using a push notification service such as Firebase Cloud Messaging or Apple Push Notification Service. Specifically, the server creates a notification message and sends it to the user's smartphone or PC.

[1314] Step 8:

[1315] The server stores the analysis results and sentiment analysis results in a database. The input is the analysis results and sentiment analysis results obtained in steps 4 and 5, and the output is the data stored in the database. This data will also be used as a training dataset in the future. Specifically, the server stores the data using an SQL database or NoSQL database.

[1316] Step 9:

[1317] The server periodically retrains the analysis device's model based on new data to improve analysis accuracy. The input is the training data stored in the database, and the output is the retrained analysis model. Specific operations include batch training and updating the model weights. Possible libraries used are TensorFlow and PyTorch.

[1318] (Application example 2)

[1319] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1320] In modern society, security threats are increasing daily, and there is a need for rapid and accurate detection and response of abnormal behavior, especially in large commercial and public facilities. While numerous surveillance cameras are installed in such locations, it is difficult to monitor all camera footage in real time due to limited human resources. Furthermore, emotion analysis is also important in detecting abnormal behavior, making it necessary to discover and respond to potential threats early. Furthermore, there is a need for a method to solve these issues while protecting individual privacy.

[1321] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1322] In this invention, the server includes means for acquiring video data from a monitoring device, means for preprocessing the acquired video data, means for transmitting the preprocessed video data to an analysis device, means for the analysis device to analyze the video data and detect anomalous behavior, means for the analysis device to analyze a person's emotions, means for anonymizing the analysis results and protecting personally identifiable information, means for notifying a user terminal and related systems of the anonymized analysis results, means for storing the analysis results in a database and using them as system learning data, means for retraining the analysis device's model based on new data to improve analysis accuracy, and means for displaying the notified analysis results on a user terminal. This enables security staff to check anomalous behavior and emotion analysis results in real time and quickly recognize and respond to potential threats.

[1323] "Monitoring equipment" refers to equipment used to acquire video data, and primarily refers to fixed cameras and dashcams.

[1324] "Video data" refers to visual information acquired from a monitoring device, and is digital data that records movements and situations.

[1325] "Preprocessing" refers to the process of performing processes such as noise removal and frame rate normalization on video data.

[1326] An "analysis device" is a device that uses preprocessed video data to analyze people's movements, abnormal behavior, and emotions.

[1327] "Abnormal behavior" refers to any unusual behavior or action, including, for example, intrusion, violent behavior, or abnormal stopping of a vehicle.

[1328] "Emotion analysis" is a technology that analyzes a person's facial expressions and tone of voice to identify specific emotions.

[1329] "Anonymization" is a process that removes personally identifiable information from the analysis results to protect the privacy of the data.

[1330] "User terminal" refers to electronic devices used to display analysis results, such as smartphones, PCs, and smart glasses.

[1331] "Related systems" refers to other systems or institutions that receive the analysis results, including police and security companies.

[1332] A "database" is a system for systematically storing and managing analysis results and learning data.

[1333] "Relearning" is a process of updating the model of the analysis device based on new data to improve the accuracy of the analysis.

[1334] "Notification" refers to the act of notifying the analysis results to the appropriate user terminal or related system.

[1335] The system according to the present invention can be implemented as a smart glasses application related to security services, and includes a monitoring device, a server, an analysis device, smart glasses, and related systems.

[1336] First, the server acquires video data in real time from monitoring devices (such as fixed cameras) connected to the network. This video data is temporarily stored in storage. The server then preprocesses the acquired video data, removing noise and normalizing the frame rate. The preprocessed video data is then sent to the analysis device.

[1337] The analyzer then receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. This analysis detects human movements and abnormal behavior (e.g., intrusions, violent behavior, and abnormal vehicle stalls). Furthermore, the analyzer incorporates an emotion engine that analyzes human facial expressions and speech tones in the video data to identify emotions. This emotion analysis can determine whether a person is displaying emotions such as fear, surprise, or anger.

[1338] Once the analysis is complete, the server retrieves the analysis results and anonymizes the data to protect privacy. Personally identifiable information is removed by methods such as facial mosaic and audio masking. Emotion analysis results are also anonymized. The server then notifies the appropriate user device and related systems of the analysis and emotion analysis results. Users receive notifications on their smart glasses and can check the results in real time. If necessary, the server also securely shares the anonymized data with relevant authorities (police, security companies).

[1339] Furthermore, the server stores the analysis results and sentiment analysis results in a database. This stored data is also used as training data for the future. The server periodically retrains the analysis device's model based on new data to improve analysis accuracy.

[1340] As a concrete example, if security staff at a large shopping mall are patrolling wearing smart glasses, images from surveillance cameras will be displayed on the glasses in real time. At the same time, the system can detect intrusions and any abnormal behavior, and the emotions of individuals who exhibit suspicious behavior (such as fear or tension) can also be displayed on the glasses. This allows security staff to quickly recognize potential threats and respond efficiently.

[1341] Here is an example of a prompt to input to a generative AI model:

[1342] Run the following code to create an application that displays abnormal behavior and emotion analysis results in real time on smart glasses worn by security staff at commercial facilities.

[1343] The surveillance equipment video analysis system detects abnormal behavior and emotions in real time. The detected information is displayed on the security staff's smart glasses. Specifically, the system performs noise removal and frame rate normalization on the video data, abnormal behavior analysis using a deep learning model, and emotion analysis using an emotion engine, and displays the results on the smart glasses.

[1344] Libraries used:

[1345] cv2 (OpenCV)

[1346] Deep learning models (e.g., DeepFace)

[1347] Smart Glasses API Library

[1348] Required functions:

[1349] get_camera_feed(): Gets video data from the server

[1350] preprocess_frame(frame): Preprocesses video data

[1351] detect_anomalies(frame): Detect abnormal behavior

[1352] analyze_emotions(frame): Analyze emotions

[1353] display_on_glasses(info): Display information on smart glasses

[1354] Please generate Python code that meets the above specifications.

[1355] As such, this system has a wide range of applications in the security field, particularly in terms of its ability to detect abnormal behavior in real time and respond quickly through emotion analysis.

[1356] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1357] Step 1:

[1358] The server acquires video data in real time from a monitoring device (such as a fixed camera). In this step, the server receives the video data sent from the monitoring device as streaming data and temporarily stores it in storage.

[1359] Step 2:

[1360] The server performs noise reduction and frame rate normalization on the acquired video data. The input is raw video data stored in storage, and the server applies noise reduction filters and frame rate unification processing to it before outputting the preprocessed video data.

[1361] Step 3:

[1362] The server transmits the preprocessed video data to the analysis device. The input is the preprocessed video data, and the server transmits this to the analysis device via the network.

[1363] Step 4:

[1364] The analysis device receives the preprocessed video data and uses a deep learning model to analyze human movements and abnormal behavior. The input of this step is the preprocessed video data, and the output is the detected abnormal behavior (e.g., intrusion, violent behavior, abnormal car stalling, etc.) along with its timestamp and location information.

[1365] Step 5:

[1366] The analyzer identifies emotions from video data by analyzing facial expressions and tone of speech. The input is pre-processed video data, and the output is the identified emotion (e.g., fear, surprise, anger, etc.). The analyzer does this using an emotion engine.

[1367] Step 6:

[1368] The server receives the analysis results and emotion analysis results and anonymizes the data to protect privacy. The input is the abnormal behavior and emotion analysis results provided by the analysis device, and the output is the anonymized analysis results. The server performs operations such as face mosaic processing and audio masking.

[1369] Step 7:

[1370] The server notifies the user device and related systems of the anonymized analysis results. The input is the anonymized analysis results, which the server notifies at the appropriate time. This allows the results to be displayed in real time on the user device (such as smart glasses).

[1371] Step 8:

[1372] The server stores the analysis results and sentiment analysis results in a database. The inputs to this step are the analysis results and sentiment analysis results, and the output is the stored data. The stored data can also be used as learning data in the future.

[1373] Step 9:

[1374] The server retrains the analysis device's model based on new data to improve analysis accuracy. The input is the newly saved data, which the server uses to retrain the deep learning model. The output is an analysis model with improved accuracy.

[1375] The detailed inputs, processing, and outputs for each step are shown below:

[1376] Step 1:

[1377] Input: Real-time video data from a surveillance device

[1378] Processing: The server receives the video data and temporarily stores it in storage.

[1379] Output: Raw video data stored in storage

[1380] Step 2:

[1381] Input: Raw video data stored in storage

[1382] Processing: Noise reduction filter applied, frame rate unified processing

[1383] Output: Pre-processed video data

[1384] Step 3:

[1385] Input: Preprocessed video data

[1386] Processing: Send data to an analysis device via a network

[1387] Output: Video data transmission to analysis device completed

[1388] Step 4:

[1389] Input: Preprocessed video data

[1390] Processing: Analyzing movement and abnormal behavior with deep learning models

[1391] Output: Detected abnormal behavior (timestamp and location information)

[1392] Step 5:

[1393] Input: Preprocessed video data

[1394] Processing: Identifying emotions with the emotion engine

[1395] Output: Identified emotion (e.g., fear, surprise, anger)

[1396] Step 6:

[1397] Input: Abnormal behavior and emotion analysis results

[1398] Processing: Anonymization by face mosaic processing, voice masking, etc.

[1399] Output: Anonymized analysis results

[1400] Step 7:

[1401] Input: Anonymized analysis results

[1402] Processing: Notify user terminals and related systems

[1403] Output: Notification to user terminal and related systems completed

[1404] Step 8:

[1405] Input: Analysis results and sentiment analysis results

[1406] Process: Save to database

[1407] Output: Data stored in the database

[1408] Step 9:

[1409] Input: Newly saved data

[1410] Processing: Retraining a deep learning model

[1411] Output: Improved analytical model

[1412] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1413] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1414] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1415] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1416] FIG. 9 illustrates an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and behaviors arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1417] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1418] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1419] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1420] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1421] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1422] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1423] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1424] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1425] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1426] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1427] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1428] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1429] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1430] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1431] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1432] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1433] The following is further disclosed regarding the above embodiment.

[1434] (Claim 1)

[1435] means for acquiring video data from a monitoring device;

[1436] means for pre-processing the acquired video data;

[1437] means for transmitting the preprocessed video data to an analysis device;

[1438] An analysis device analyzes the video data and detects abnormal behavior;

[1439] Measures to anonymize the analysis results and protect personally identifiable information;

[1440] a means for notifying the user terminal and related systems of the anonymized analysis results;

[1441] A means of storing the analysis results in a database and using them as training data for the system;

[1442] A system including a means for retraining the model of the analysis device based on new data to improve analysis accuracy.

[1443] (Claim 2)

[1444] 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

[1445] (Claim 3)

[1446] 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device.

[1447] "Example 1"

[1448] (Claim 1)

[1449] means for acquiring video data from a monitoring device;

[1450] means for pre-processing the acquired video data;

[1451] means for transmitting the preprocessed video data to an analysis device;

[1452] An analysis device analyzes the video data and detects abnormal behavior;

[1453] Measures to anonymize the analysis results and protect personally identifiable information;

[1454] a means for notifying the user terminal and related systems of the anonymized analysis results;

[1455] A means of storing the analysis results in a database and using them as training data for the system;

[1456] A system including a means for retraining the model of the analysis device based on new data to improve analysis accuracy.

[1457] (Claim 2)

[1458] 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

[1459] (Claim 3)

[1460] 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device.

[1461] (Claim 4)

[1462] 10. The system of claim 1, further comprising means for performing face mosaic processing and voice masking when obtaining the analysis results and anonymizing the data for privacy protection.

[1463] (Claim 5)

[1464] The system according to claim 1, further comprising means for using an API or a push notification system when notifying the user terminal of the analysis results.

[1465] (Claim 6)

[1466] 10. The system of claim 1, further comprising means for acquiring, pre-processing and transferring video data from the monitoring device in real time to an analysis device.

[1467] "Application Example 1"

[1468] (Claim 1)

[1469] means for acquiring video data from a monitoring device;

[1470] means for pre-processing the acquired video data;

[1471] means for transmitting the preprocessed video data to an analysis device;

[1472] An analysis device analyzes the video data and detects abnormal behavior;

[1473] Measures to anonymize the analysis results and protect personally identifiable information;

[1474] a means for notifying the user terminal and related systems of the anonymized analysis results;

[1475] A means of storing the analysis results in a database and using them as training data for the system;

[1476] A means for retraining the model of the analysis device based on new data to improve the accuracy of the analysis;

[1477] A means of sending real-time abnormal behavior alerts to a smartphone application;

[1478] A system including a means for allowing a user to view past abnormal behavior detection history.

[1479] (Claim 2)

[1480] 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

[1481] (Claim 3)

[1482] 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device.

[1483] "Example 2: Combining Emotion Engines"

[1484] (Claim 1)

[1485] means for acquiring video data from a monitoring device;

[1486] means for pre-processing the acquired video data;

[1487] means for transmitting the preprocessed video data to an analysis device;

[1488] An analysis device analyzes the video data and detects abnormal behavior;

[1489] A means for identifying a user's emotion from video data;

[1490] Measures to anonymize the analysis results and protect personally identifiable information;

[1491] a means for notifying the user terminal and related systems of the anonymized analysis results;

[1492] A means of storing the analysis results in a database and using them as training data for the system;

[1493] A system including a means for retraining the model of the analysis device based on new data to improve analysis accuracy.

[1494] (Claim 2)

[1495] 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

[1496] (Claim 3)

[1497] 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device.

[1498] "Application example 2 when combining emotion engines"

[1499] (Claim 1)

[1500] means for acquiring video data from a monitoring device;

[1501] means for pre-processing the acquired video data;

[1502] means for transmitting the preprocessed video data to an analysis device;

[1503] An analysis device analyzes the video data and detects abnormal behavior;

[1504] A means for the analysis device to analyze the emotions of a person;

[1505] Measures to anonymize the analysis results and protect personally identifiable information;

[1506] a means for notifying the user terminal and related systems of the anonymized analysis results;

[1507] A means of storing the analysis results in a database and using them as training data for the system;

[1508] A means for retraining the model of the analysis device based on new data to improve the accuracy of the analysis;

[1509] means for displaying the notified analysis results on a user terminal;

[1510] A system including:

[1511] (Claim 2)

[1512] 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

[1513] (Claim 3)

[1514] 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device. [Explanation of symbols]

[1515] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for acquiring video data from a monitoring device; means for pre-processing the acquired video data; means for transmitting the preprocessed video data to an analysis device; An analysis device analyzes the video data and detects abnormal behavior; Measures to anonymize the analysis results and protect personally identifiable information; a means for notifying the user terminal and related systems of the anonymized analysis results; A means of storing the analysis results in a database and using them as training data for the system; A system including a means for retraining the model of the analysis device based on new data to improve analysis accuracy.

2. 10. The system of claim 1, further comprising means for performing noise reduction and frame rate normalization on the preprocessed video data.

3. 10. The system of claim 1, further comprising means for recording timestamps and location information of abnormal behavior detected by the analysis device.

Images