system

Abstract

We provide the system. [Solution] A means of detecting incoming calls from unregistered callers, A means for acquiring the audio of an incoming call and processing the audio data in real time, A means of converting acquired audio into text information and comparing it with fraudulent talk patterns, A means to determine the possibility of fraudulent activity, start recording and notify a warning if a threshold is exceeded, A means of quickly displaying the results on the user's device, A system that includes a means of generating and outputting a warning voice message to the caller if fraudulent activity is detected.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot, which is performed by at least one processor, the method including steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a character of the chatbot, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance as a response to the user utterance.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The problem of fraud calls, which still causes serious damage especially to the elderly in society, is not sufficient with only conventional warnings, and effective preventive measures against sophisticated fraud talks are required. The present invention aims to detect fraud calls in real time and prevent damage. However, since fraud methods evolve frequently, it is necessary to cope with new patterns each time.

Means for Solving the Problems

[0005] This invention employs a means to automatically detect incoming calls from unregistered callers and convert the call audio into text in real time. The converted text is then compared with past fraudulent talk patterns to determine the likelihood of fraud. If the likelihood exceeds a predetermined threshold, the system immediately starts recording the call and notifies both parties involved in the call of a warning. Furthermore, it has the functionality to periodically update the database of fraudulent talk patterns and adapt to new fraudulent methods using an AI learning model. As a result, users can gain peace of mind by reviewing the recorded data and the judgment results.

[0006] "Unregistered caller" refers to a caller's phone number that is not included in the user's pre-registered phone number list.

[0007] "Means for detecting incoming calls" refers to systems or mechanisms that automatically identify incoming calls from unregistered callers and activate certain triggers.

[0008] "Means for processing audio data" refers to methods and devices for analyzing audio generated during a call in real time and converting it into a format suitable for subsequent processing.

[0009] "Means of converting to text" refers to methods and devices that convert audio data into corresponding text information using technologies such as speech recognition.

[0010] A "fraudulent talk pattern" refers to a collection of characteristic phrases and expressions extracted from previously recorded fraud-related phone calls.

[0011] A "threshold" is a reference value used to determine the likelihood of fraud; it is a numerical value that, if exceeded, triggers a specific action.

[0012] "Means for initiating recording" refers to a system or method that automatically starts the process of recording the content of a phone call based on certain criteria.

[0013] A "warning notification system" is a means or mechanism for providing audio or visual alerts in response to detected potential fraud.

[0014] "Updating a database" refers to the process of reflecting newly acquired information or patterns in existing records to maintain an up-to-date state.

[0015] A "learning model" refers to the algorithms and structures that artificial intelligence uses to perform pattern recognition and prediction of new information based on past data. [Brief explanation of the drawing]

[0016] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Figure 11]It is a sequence diagram showing the processing flow of the data processing system in Embodiment 1. [Figure 12] It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Embodiment 2 when the emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when the emotion engine is combined.

Mode for Carrying Out the Invention

[0017] Hereinafter, an example of an embodiment of the system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, the numbered processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Also, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), and the like.

[0020] In the following embodiments, the numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0021] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0022] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0024] [First Embodiment]

[0025] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0026] As shown in Figure 1, the 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.

[0027] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0029] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and 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.

[0030] 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 perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0031] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0032] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] As shown in Figure 2, in the data processing device 12, a specific processing 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" related to the technology of this 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 according to the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0036] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0037] This invention is a fraud prevention system that monitors incoming calls from unregistered callers in real time and determines the possibility of fraud, thereby preventing damage. The operation of the entire system is described below.

[0038] The server constantly monitors incoming calls to the user's device. If a call comes in from a number other than those registered by the user, it automatically initiates an AI-based analysis process.

[0039] The terminal collects audio data as soon as a call is received and streams it to the server. The transmitted data is converted to text using speech recognition technology, making the call content available for analysis.

[0040] The server converts the audio to text and then compares the text against a database containing fraudulent talk patterns. This process applies a pattern recognition algorithm based on past fraud cases, quantifying the likelihood of fraud.

[0041] When the likelihood of fraud exceeds a certain threshold, the server initiates a recording process. This allows the call to be stored as evidence. Simultaneously, a warning message is generated in audio form and notified to both parties involved in the call, alerting the recipient and providing a psychological deterrent to the caller.

[0042] After the call, the server notifies the user of the details of the fraud detection and provides the recording data. This feature is designed to allow users to review the call content later, providing them with peace of mind.

[0043] The servers regularly update their fraud database and use AI learning models to flexibly adapt to new fraudulent methods. This automated update process ensures the system is always up-to-date, protecting users from fraud over the long term.

[0044] As a concrete example, consider a scenario where a caller attempts to commit a wire fraud by impersonating the user's "son." The server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. As a result, recording begins immediately, and the user is notified with a warning. This process effectively reduces the risk of the user being deceived by a fraudulent call.

[0045] The following describes the processing flow.

[0046] Step 1:

[0047] The server monitors all incoming calls to the user's device and detects incoming calls from unregistered callers. This detection triggers the start of AI processing.

[0048] Step 2:

[0049] As soon as a call begins, the terminal collects voice data in real time and streams it to the server. During this process, the voice data is compressed to optimize bandwidth usage.

[0050] Step 3:

[0051] The server converts the received audio data into text using speech recognition technology. It also performs noise reduction processing to improve audio clarity.

[0052] Step 4:

[0053] The server compares the converted text against a database of scam talk patterns. Using natural language processing algorithms, it compares extracted key phrases and performs pattern recognition.

[0054] Step 5:

[0055] The server scores the likelihood of fraud. If the calculated score exceeds a threshold, it determines that there is a high probability of fraud and proceeds to the next step.

[0056] Step 6:

[0057] The server starts recording as soon as it determines that a call is highly likely to be a scam. This recording is saved as evidence and remains accessible to the user even after the call ends.

[0058] Step 7:

[0059] The server uses speech synthesis technology to generate a warning message and notifies both parties in the call in real time. This draws the listener's attention and puts pressure on the caller.

[0060] Step 8:

[0061] After the call ends, the server notifies the user of the fraud detection result and guides them to view the recorded data on a dedicated portal or application.

[0062] Step 9:

[0063] The server updates the fraudulent talk pattern database based on call data and judgment results. The AI ​​model continuously learns using newly collected data to improve its accuracy.

[0064] (Example 1)

[0065] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0066] In modern times, telephone fraud is becoming increasingly sophisticated year by year. This raises the risk of ordinary consumers unintentionally becoming victims of fraud, posing a significant security challenge, especially for the elderly and individuals with low IT literacy. Traditional methods such as manual registration restrictions and simple suspicious number blocking are insufficient; therefore, a comprehensive fraud prevention system employing new methods is needed.

[0067] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0068] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and streaming the audio data in real time using a data processing device, and means for converting the captured audio into text using speech recognition technology and comparing it with fraudulent talk patterns. This enables automatic detection of fraudulent activity and immediate response to it.

[0069] "Unregistered caller ID" refers to phone numbers or contact information that the user has not previously registered in the system.

[0070] "Means for detecting incoming calls" refers to a device or method that monitors telephone communications and identifies calls from unregistered callers.

[0071] "Means for capturing audio" refers to a device or mechanism for capturing audio signals generated during a telephone call and inputting them into a system.

[0072] A "data processing device" refers to a hardware or software component used for processing, converting, and analyzing audio signals.

[0073] "Means for streaming audio data" refers to a technology or method for transmitting acquired audio data in real time to other devices or systems.

[0074] "Speech recognition technology" refers to the technology that receives speech signals and converts them into text format that can be understood as human language.

[0075] "Fraudulent talk patterns" refer to typical conversational content and phrase patterns from past fraud cases, and are used as criteria for diagnosing the possibility of fraud.

[0076] A "pattern recognition algorithm" refers to a program or method that identifies certain regularities or features in data and performs classification or judgment based on specific conditions.

[0077] A "generative AI model" refers to an artificial intelligence model that learns new data patterns and uses them to generate data and make inferences.

[0078] By using the following configuration and technology in the implementation of this fraud prevention system, it is possible to smoothly and efficiently detect potential fraud and protect users.

[0079] First, the device receives an incoming call through the user's phone function. Upon receiving the call, the device's microphone is activated, and audio data from the call is collected in real time. This audio data is immediately streamed to the server via the communication network. Real-time data processing technologies such as WebRTC are used for streaming.

[0080] The server uses the received audio data to convert the speech to text using speech recognition software (e.g., Google® Cloud Speech-to-Text). The transcribed data is then compared against a database containing recorded fraudulent talk patterns. Here, a pattern recognition algorithm using a natural language processing library (e.g., NLTK or Spacy) is applied to quantify the likelihood of fraud.

[0081] If a call is deemed highly likely to be fraudulent, the server immediately initiates a recording process. This stores the call content as evidence and generates a warning message for the user. This message generation uses synthesized speech technology (e.g., Amazon Polly). The warning is sent directly to both the caller and the recipient in real time during the call.

[0082] After the call, the server notifies the user of the details of the fraud assessment and the recorded data, prompting them to review it. The notification is sent via email or application push notification, and the user can review the content on a secure portal site or dedicated application. This information allows the user to review the call content and take steps to prevent becoming a victim of fraud.

[0083] Furthermore, the server automatically and regularly updates its database of fraudulent talk patterns and continuously learns to counter new fraudulent tactics using its generating AI model. This ensures that the system always reflects the latest fraud information and maintains a form that can adapt to ever-evolving fraudulent methods.

[0084] For example, if a call comes in from someone impersonating a "son" attempting to commit a wire transfer scam, the server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. At this stage, recording begins and a warning is sent to the user.

[0085] Example of a prompt:

[0086] "Is this call potentially a scam? The keywords included are 'accident' and 'urgent transfer'."

[0087] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0088] Step 1:

[0089] When a call is received, the terminal activates its built-in microphone and collects the call audio in real time. This audio data is the input. In other words, the terminal converts the audio signal into a digital format and sends it to the server in the appropriate format. The audio data, as output, is then moved to the next processing stage using real-time transmission technology.

[0090] Step 2:

[0091] The server receives audio data sent from the terminal. Next, a speech recognition component converts this audio data into text. The input data is an audio signal, and the output data is text in string format. Speech recognition technology is used to extract the content of the audio as textual information.

[0092] Step 3:

[0093] The server uses the converted text data to compare it against a database containing registered fraudulent talk patterns. Here, a pattern recognition algorithm is used to detect characteristic phrases of fraud and determine the likelihood of it being a scam. The input data is text, and the output data is a numerical rating of the likelihood of fraud.

[0094] Step 4:

[0095] If the likelihood of fraud exceeds a certain threshold, the server initiates the recording process. The input is the result of the fraud detection. As output, the recorded audio data is saved to the storage system. Simultaneously, a warning message is generated and sent to the user and caller using synthesized speech technology. This serves to alert the recipient.

[0096] Step 5:

[0097] After the call ends, the server notifies the user of the judgment result and the recorded data. The input is the recorded data and the judgment result, and the output is the notification information to the user. The notification is sent via email or application, and the user can check the content on a dedicated portal site or application.

[0098] Step 6:

[0099] The server automatically and regularly updates its database of fraudulent talk patterns and uses a generating AI model to flexibly adapt to new fraudulent methods. The input to this process is updated fraud information and historical data trends, while the output is an updated database and an improved model. In this way, the system can always operate with the most up-to-date fraud information.

[0100] (Application Example 1)

[0101] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0102] With the advancement of communication technology, fraudulent activities have diversified, and scams exploiting voice calls are on the rise. Such fraudulent activities are a serious problem, making it difficult for users to use voice calls with peace of mind. In particular, vulnerable groups such as the elderly are at risk of suffering significant losses due to the lack of adequate countermeasures against fraud. In this situation, there is a need for an efficient system that prevents fraudulent activities and provides users with a sense of security.

[0103] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0104] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for acquiring the audio of the incoming communication and processing the audio data in real time, and means for converting the acquired audio into text information and comparing it with fraudulent talk patterns. This enables the automatic detection and warning of potential fraudulent activity, allowing users to use voice calls with peace of mind.

[0105] "Incoming caller" refers to the identification information of the party that initiated the call or message in a communication method.

[0106] "Audio data" refers to information that represents sound in digital or analog format.

[0107] "Textual information" refers to text data obtained by converting data such as audio into a string of characters.

[0108] "Fraudulent behavior talk patterns" are a collection of characteristic voices and phrases extracted from past cases of fraudulent activity.

[0109] "Recording" is the process of saving certain data or information to prepare for future use.

[0110] A "warning" is a message issued to draw attention to risks or potential dangers.

[0111] A "warning audio message" is an audio notification used to warn or caution the sender or recipient when fraudulent activity is suspected.

[0112] A "user" is an entity that operates or utilizes a system, product, or service.

[0113] A "dedicated information terminal" is an electronic device designed specifically for a particular purpose or function.

[0114] An "application program" is software developed to achieve a specific function or purpose.

[0115] The system realizing this invention mainly consists of three entities: a server, a terminal, and a user. The server is responsible for monitoring calls from unregistered incoming sources and processing those calls in real time. The terminal acquires the audio data of the call and streams it to the server. The server receives this and converts the audio into text data using the Google Cloud Speech-to-Text API. Based on this text data, it is compared against fraudulent talk patterns, and an AI model using TENSORFLOW® or PyTorch analyzes and determines the possibility of fraud.

[0116] If the server detects a potential for fraud exceeding a threshold, the call is recorded, and a warning message is displayed on the user's device. Additionally, a cautionary voice message is generated and sent to the caller. This process allows the user to continue the call with confidence and later review the call content.

[0117] As a concrete example, consider a scenario where a user receives a phone call saying, "You've won a prize, please provide your bank account information." This system detects and analyzes phrases like "prize" and "bank account information," and if it determines that there is a high probability of fraudulent activity, it immediately starts recording and warns the user that "this may be a scam." An example of a prompt message would be, "Analyze the content of the call received by the user, analyze whether it contains the key phrases 'prize' and 'bank account information,' and issue a warning if it is suspicious."

[0118] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0119] Step 1:

[0120] The terminal detects an incoming call from an unregistered source and acquires audio data as soon as the call begins. This audio data serves as the input and is streamed to the server in real time.

[0121] Step 2:

[0122] The server converts the received audio data into text information using the Google Cloud Speech-to-Text API. This process outputs the audio data as text data.

[0123] Step 3:

[0124] The server compares the generated text data with a database of fraudulent talk patterns. A generative AI model using TensorFlow or PyTorch performs a probabilistic determination by comparing it with a database based on past cases. This process outputs a numerical result indicating the likelihood of fraudulent activity.

[0125] Step 4:

[0126] The server determines a threshold based on the quantified result, and if the likelihood of fraudulent activity exceeds the threshold, it starts recording the call. The input here is the judgment result, and the output is the start of the recording process.

[0127] Step 5:

[0128] A warning message based on the judgment result is displayed on the user's terminal. Warning information from the server is input and output to the user in the form of a notification.

[0129] Step 6:

[0130] The server generates a warning voice message and sends it to the caller. This process outputs a generated voice file, which is then played back by the caller to deliver the warning.

[0131] Step 7:

[0132] After the call ends, the server sends the judgment result and recorded data to the user's dedicated information terminal or application program for user review. In this step, the recorded data is the input, and it is output by being displayed on the user's review interface.

[0133] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0134] This invention combines an emotion engine with a fraud prevention system, providing a function to detect the user's emotions in real time during a call and more accurately determine the likelihood of fraud. This system monitors incoming calls from unregistered callers and analyzes the call content and the user's emotions to improve its ability to respond to fraud.

[0135] When a call comes in to a user's device, if it's from an unregistered caller, the server activates the emotion engine simultaneously with audio collection. Using speech recognition technology, it converts the audio into text and analyzes the user's emotional state from their voice patterns while comparing it against fraudulent talk patterns.

[0136] The device captures audio during a call with high accuracy and sends the necessary data to the server for the emotion engine. Factors such as tone, speed, and emphasis are considered.

[0137] The server uses an emotion engine to detect the user's emotions. The detected emotion data is fed back into the fraud possibility assessment and contributes to an overall decision. This result influences the fraud threshold determination and, if necessary, initiates recording or issues a warning.

[0138] If a user is emotionally distressed, the server will detect this and adjust the content and timing of the warning notification to help the user understand. For example, an additional warning in a calmer tone may be issued.

[0139] After the call ends, the server notifies the user of the detected emotion data and analysis results. This allows the user to understand their own emotional responses and use that information to improve future interactions.

[0140] Emotional data is stored as part of an ongoing learning process, contributing to the updating of the fraudulent talk pattern database and improving the accuracy of the AI ​​model. Using this learning process, the system continues to evolve and adapt to new fraudulent tactics and user emotional patterns.

[0141] For example, if someone claiming to be from a "financial institution" requests "account information verification," and the server detects that the user's voice indicates distress, the server will intensify the warning and provide an interaction prompting the user to acknowledge their emotions and exercise caution.

[0142] The following describes the processing flow.

[0143] Step 1:

[0144] The server monitors incoming calls to the user's terminal and detects calls from unregistered callers. This detection triggers the activation of voice data processing and the emotion engine.

[0145] Step 2:

[0146] As soon as a call starts, the device captures voice data with high precision and sends it to the server in stream format, including the voice parameters necessary for the emotion engine.

[0147] Step 3:

[0148] The server uses speech recognition technology to convert the audio data into text and compares it against a database of fraudulent talk patterns. In parallel, an emotion engine analyzes the tone, speed, emphasis, and other factors of the voice.

[0149] Step 4:

[0150] The server receives analysis results from the emotion engine and provides feedback to the fraud detection score if the user's emotions indicate anxiety or distress.

[0151] Step 5:

[0152] If the fraud probability score exceeds a threshold, the server will start recording and notify the user with a warning message tailored to their emotional state.

[0153] Step 6:

[0154] If a user shows emotional changes during a call, the server detects this in real time and dynamically adjusts the content and intensity of warning notifications to support user understanding.

[0155] Step 7:

[0156] After the call ends, the server notifies the user of the results, audio data, and sentiment analysis results. The user can then view this information through a dedicated portal or application.

[0157] Step 8:

[0158] The server updates the database using sentiment data and call history, and the AI ​​model continues to learn from the new data. This process allows the system to evolve and improve the accuracy of future fraud detections.

[0159] (Example 2)

[0160] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0161] Traditional fraud prevention systems relied on standard voice recognition and simple pattern matching for communications from unregistered sources, making it difficult to accurately determine the likelihood of fraud. Furthermore, they could not take into account the user's emotional state, resulting in inappropriate timing and content of warnings regarding fraudulent methods. Additionally, they struggled to respond quickly to the emergence of new fraud techniques.

[0162] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0163] In this invention, the server includes means for capturing transmitted audio and processing the audio information in real time, means for converting the captured audio into text and comparing it with fraud patterns, and means for analyzing the characteristics of the audio and determining the emotional state. This makes it possible to determine the possibility of fraud with high accuracy while taking into account the user's emotional state and to provide appropriate warnings and advice in real time. Furthermore, by regularly updating the information source for fraudulent talk patterns and improving accuracy through a learning device, it is possible to respond quickly to new fraudulent methods.

[0164] "Means for detecting communications" refers to a function that checks for and detects the presence of calls and messages from unregistered callers.

[0165] "Means for processing voice information" refers to a function that receives voice during a call and converts and manages it as digital data in real time.

[0166] "Means of converting to text" refers to a function that performs the process of converting audio data into text information using speech recognition technology.

[0167] "Methods for matching with fraud patterns" refers to a function that compares transcribed audio information with known fraudulent talk patterns and analyzes the degree of similarity.

[0168] "A means of analyzing voice characteristics and determining emotional state" refers to a function that analyzes features such as tone, speed, and emphasis of voice to identify the user's emotional response.

[0169] "Means of providing warnings and advice" refers to a function that alerts users and provides specific countermeasures when a potential scam is detected.

[0170] "Methods for regularly updating information sources" refers to the process of keeping fraud databases up-to-date and incorporating new fraud patterns.

[0171] A "learning device" is a technology necessary to train an AI model based on new information and continuously improve the system's discrimination accuracy.

[0172] This invention is a system aimed at preventing fraud, providing a function to analyze the user's emotions in real time during communication and determine the possibility of fraud. To achieve this, a system configuration in which a server and terminals are linked is necessary.

[0173] The server receives communications from the user's terminal. Specific software used includes speech recognition technology and generative AI models. The server transcribes the received audio data into text in real time and compares it against a fraud pattern database. During this process, features extracted from the audio data are used to activate an emotion engine and analyze the user's emotional state.

[0174] The terminal is responsible for capturing audio during calls with high precision. Hardware components include a microphone and an audio processing unit. The terminal converts the audio signal into digital data and sends it to the server. This allows the server to perform more accurate analysis.

[0175] As a concrete example, consider a scenario where someone claiming to be from a "financial institution" requests "confirmation of account information" from an unknown caller. In this situation, where the user is inevitably suspicious of fraud, the system detects the user's emotional distress and immediately intensifies the warning. The user is given calm voice advice such as, "Stay calm and do not carelessly provide personal information." This process is achieved by instructing the system to "detect keywords suggesting potential fraud through speech recognition and also perform emotional analysis of the user's emotional state."

[0176] In this system, emotional data is accumulated through a continuous learning process, and the fraudulent talk pattern database is also updated regularly. This allows the system to continuously adapt to new fraudulent tactics and changing user emotional patterns. By receiving this feedback, users can better understand their future unconscious reactions and take countermeasures.

[0177] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0178] Step 1:

[0179] The terminal detects incoming calls. Inputs include the caller's number and the call start signal. If the call is from an unregistered caller, the terminal immediately begins capturing audio data. The audio is captured by the microphone, noise is removed to produce clear audio data, and it is converted to a digital format for output.

[0180] Step 2:

[0181] The terminal transmits the captured audio data to the server in real time. During this process, acoustic characteristics such as tone, speed, and emphasis are input and sent directly to the server as digital data.

[0182] Step 3:

[0183] The server performs speech recognition based on the received audio data. Using a generative AI model, it converts speech to text, outputting text data from the audio data as input. During this process, advanced algorithms are employed to perform precise analysis of the audio data.

[0184] Step 4:

[0185] The server compares the transcribed text against a database of fraud patterns. It compares the text data received as input against known fraud patterns and outputs a degree of similarity based on the results. Based on the matching results, it makes an initial determination of the likelihood of fraud.

[0186] Step 5:

[0187] The server activates an emotion engine to analyze the user's emotions from the voice data. It takes voice characteristic data as input and outputs an emotional state (e.g., calm, agitated). The emotion analysis results are fed back into the fraud detection process.

[0188] Step 6:

[0189] The server comprehensively assesses the likelihood of fraud, issues a warning if necessary, and begins recording. In this step, it uses whether the assessed fraud likelihood score exceeds a threshold as a criterion, and adjusts the content and timing of the warning notification before sending it to the user. The output includes the warning content and stored recording data.

[0190] Step 7:

[0191] The server provides emotional feedback to the user after the call ends. Using the analyzed emotional data and fraud detection results as input, it outputs analysis results as appropriate and notifies the user via their terminal or a dedicated digital platform. Based on this, the user can consider future countermeasures.

[0192] Step 8:

[0193] The server feeds back the collected data to the learning device, updating the generated AI model. This process involves taking new fraud patterns and emotional patterns as input and outputting improvements in AI accuracy and system evolution.

[0194] (Application Example 2)

[0195] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0196] In recent years, telephone fraud tactics have become increasingly sophisticated, with a particular emphasis on appealing to emotions. Traditional fraud prevention systems rely solely on matching voice patterns and keywords, failing to consider the user's emotional state, which can lead to false positives and user confusion. Furthermore, users rarely have the opportunity to recognize their own emotional responses during a call, making it difficult for them to take preventative measures.

[0197] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0198] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and processing the audio data in real time, and means for analyzing the user's emotions from their voice patterns and determining the likelihood of fraud based on the emotional data. This makes it possible to more accurately assess the risk of fraud while recognizing the user's emotional response and to take warnings and risk avoidance actions.

[0199] "Unregistered caller" refers to an incoming call from a phone number that is not pre-registered on the user's device or system.

[0200] "Audio data processing" refers to a series of operations that analyze or transform audio acquired during a call for a specific purpose.

[0201] "Converting speech to text" is the process of automatically replacing spoken information generated during a phone call with written text.

[0202] A "fraudulent talk pattern" refers to the characteristics of audio data that include conversational content or specific phrases intended for fraudulent purposes.

[0203] "Emotional data" refers to information that quantifies or represents the emotional tendencies and reactions that a user exhibits during a call.

[0204] "Notifying a warning" refers to providing users with information to inform them of the risks and encourage them to take safety measures when fraud is suspected.

[0205] A "learning model" is an algorithm or system that uses machine learning to recognize and predict specific data patterns.

[0206] "Emotional feedback" is an information provision process that informs users about the content and impact of their emotional responses during a call.

[0207] This invention is a system that uses the user's communication device to analyze the possibility of fraud in real time on the server side and issue warnings based on sentiment data. To implement this system, the user's terminal first detects an incoming call from an unregistered source. When a voice call begins, the terminal captures the audio during the call with high accuracy and transmits it to the server in real time.

[0208] The server uses speech recognition technologies such as the Google Cloud Speech-to-Text API to convert received audio into text data. This text data is compared against a database of fraudulent talk patterns and analyzed to determine the likelihood of fraud. Simultaneously, the server uses the Python transformers library to analyze the user's emotional state from the transcribed conversation. The emotional data obtained in this process is fed back into the judgment process.

[0209] If the server determines that a user is facing a risk of fraud, it sends a warning notification to the user's device via the Flutter® framework, and the audio is recorded. This allows the user to understand the risks during the call and take appropriate action. After the call ends, the server also provides the user with emotional feedback and encourages self-analysis of past calls through a dedicated application.

[0210] For example, if a user is approached with a fraudulent financial transaction offer, and the audio matches a scammer's conversation pattern, and the system detects signs of distress in the user, it will immediately issue a warning and begin recording. This allows the user to gain knowledge about the fraudulent activity and strengthen their preparations for the next steps.

[0211] An example of a prompt to a generative AI model is: "Assess the likelihood of fraud based on the following sentiment data and text in response to the caller's call, and output the result. Sentiment data: agitated, frightened. Text: 'Please provide your account information.'" By using this prompt, the AI ​​model can gain a deeper understanding of the caller's intent and effectively assess the risk of fraud.

[0212] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0213] Step 1:

[0214] The terminal detects incoming calls from unregistered callers. At this time, it receives the caller's number information as input, and if it does not exist in the internal registration database, it determines the caller is unknown and activates the system.

[0215] Step 2:

[0216] The terminal captures audio during a call and sends it to the server with high accuracy. It receives audio data as input, encodes it, and sends it to the server in real time. Here, noise cancellation technology is used to maintain clear audio data.

[0217] Step 3:

[0218] The server uses the Google Cloud Speech-to-Text API to convert audio data into text. The server receives audio data as input and applies speech recognition technology to generate text output. This output text forms the basis for the next analysis step.

[0219] Step 4:

[0220] The server uses text data to match it against scam talk patterns. Here, it takes text data as input and compares it with a scam database to perform data calculations that identify phrases that match predefined scam patterns.

[0221] Step 5:

[0222] The server uses the transformers library to analyze the user's emotional state from the text. This process quantifies the emotional data extracted from the input text and classifies it into categories such as positive, negative, and neutral. The calculated emotional data is then used in the next decision step.

[0223] Step 6:

[0224] The server determines the risk of fraud based on the degree of match with fraud patterns and the user's sentiment data. It receives sentiment data and fraud match scores as input, compares them to a threshold, and outputs whether a warning is necessary. Based on this result, it decides whether a warning is sent to the user.

[0225] Step 7:

[0226] If the server determines that the level has exceeded the danger threshold, it sends a warning notification to the terminal and starts recording. At this point, it uses the final risk assessment result as input to generate a warning message, displays the warning on the terminal, and starts the process of saving the audio data.

[0227] Step 8:

[0228] After the call ends, the server generates emotional feedback and notifies the user. The emotional data collected during the call is aggregated and converted into a feedback format that is easy for the user to understand. This result is then communicated to the user and used to improve future fraud prevention efforts.

[0229] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating 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.

[0230] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0231] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0232] [Second Embodiment]

[0233] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0234] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0235] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0236] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0237] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0238] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0239] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0240] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0241] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0242] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0243] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0244] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. 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".

[0245] This invention is a fraud prevention system that monitors incoming calls from unregistered callers in real time and determines the possibility of fraud, thereby preventing damage. The operation of the entire system is described below.

[0246] The server constantly monitors incoming calls to the user's device. If a call comes in from a number other than those registered by the user, it automatically initiates an AI-based analysis process.

[0247] The terminal collects audio data as soon as a call is received and streams it to the server. The transmitted data is converted to text using speech recognition technology, making the call content available for analysis.

[0248] The server converts the audio to text and then compares the text against a database containing fraudulent talk patterns. This process applies a pattern recognition algorithm based on past fraud cases, quantifying the likelihood of fraud.

[0249] When the likelihood of fraud exceeds a certain threshold, the server initiates a recording process. This allows the call to be stored as evidence. Simultaneously, a warning message is generated in audio form and notified to both parties involved in the call, alerting the recipient and providing a psychological deterrent to the caller.

[0250] After the call, the server notifies the user of the details of the fraud detection and provides the recording data. This feature is designed to allow users to review the call content later, providing them with peace of mind.

[0251] The servers regularly update their fraud database and use AI learning models to flexibly adapt to new fraudulent methods. This automated update process ensures the system is always up-to-date, protecting users from fraud over the long term.

[0252] As a concrete example, consider a scenario where a caller attempts to commit a wire fraud by impersonating the user's "son." The server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. As a result, recording begins immediately, and the user is notified with a warning. This process effectively reduces the risk of the user being deceived by a fraudulent call.

[0253] The following describes the processing flow.

[0254] Step 1:

[0255] The server monitors all incoming calls to the user's device and detects incoming calls from unregistered callers. This detection triggers the start of AI processing.

[0256] Step 2:

[0257] As soon as a call begins, the terminal collects voice data in real time and streams it to the server. During this process, the voice data is compressed to optimize bandwidth usage.

[0258] Step 3:

[0259] The server converts the received audio data into text using speech recognition technology. It also performs noise reduction processing to improve audio clarity.

[0260] Step 4:

[0261] The server compares the converted text against a database of scam talk patterns. Using natural language processing algorithms, it compares extracted key phrases and performs pattern recognition.

[0262] Step 5:

[0263] The server scores the likelihood of fraud. If the calculated score exceeds a threshold, it determines that there is a high probability of fraud and proceeds to the next step.

[0264] Step 6:

[0265] The server starts recording as soon as it determines that a call is highly likely to be a scam. This recording is saved as evidence and remains accessible to the user even after the call ends.

[0266] Step 7:

[0267] The server uses speech synthesis technology to generate a warning message and notifies both parties in the call in real time. This draws the listener's attention and puts pressure on the caller.

[0268] Step 8:

[0269] After the call ends, the server notifies the user of the fraud detection result and guides them to view the recorded data on a dedicated portal or application.

[0270] Step 9:

[0271] The server updates the fraudulent talk pattern database based on call data and judgment results. The AI ​​model continuously learns using newly collected data to improve its accuracy.

[0272] (Example 1)

[0273] Next, we will describe Example 1. 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".

[0274] In modern times, telephone fraud is becoming increasingly sophisticated year by year. This raises the risk of ordinary consumers unintentionally becoming victims of fraud, posing a significant security challenge, especially for the elderly and individuals with low IT literacy. Traditional methods such as manual registration restrictions and simple suspicious number blocking are insufficient; therefore, a comprehensive fraud prevention system employing new methods is needed.

[0275] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0276] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and streaming the audio data in real time using a data processing device, and means for converting the captured audio into text using speech recognition technology and comparing it with fraudulent talk patterns. This enables automatic detection of fraudulent activity and immediate response to it.

[0277] "Unregistered caller" refers to a phone number or contact that the user has not previously registered in the system.

[0278] "Means for detecting incoming calls" refers to a device or method that monitors telephone communications and identifies calls from unregistered callers.

[0279] "Means for capturing voice" refers to a device or mechanism for capturing voice signals generated during a telephone call and inputting them into the system.

[0280] "Data processing device" refers to a hardware or software component for processing, converting, and analyzing voice signals.

[0281] "Means for streaming voice data" refers to a technology or method for transmitting the acquired voice data in real time to other devices or systems.

[0282] "Speech recognition technology" refers to a technology that receives a voice signal and converts it into a text format that can be understood as human language.

[0283] "Fraudulent conversation pattern" refers to the typical conversation content and phrase patterns in past fraud incidents, which are used as criteria for diagnosing the possibility of fraud.

[0284] "Pattern recognition algorithm" refers to a program or method that identifies certain regularities and features in data and performs classification and determination based on specific conditions.

[0285] "Generative AI model" refers to an artificial intelligence model that learns new data patterns and generates or makes inferences based on them.

[0286] In the implementation of this fraud call countermeasure system, by using the following configurations and technologies, it is possible to smoothly and efficiently detect the possibility of fraud and protect users.

[0287] First, the device receives an incoming call through the user's phone function. Upon receiving the call, the device's microphone is activated, and audio data from the call is collected in real time. This audio data is immediately streamed to the server via the communication network. Real-time data processing technologies such as WebRTC are used for streaming.

[0288] The server uses the received audio data to convert it to text using speech recognition software (e.g., Google Cloud Speech-to-Text). The transcribed data is then compared against a database containing recorded fraudulent talk patterns. Here, a pattern recognition algorithm using a natural language processing library (e.g., NLTK or Spacy) is applied to quantify the likelihood of fraud.

[0289] If a call is deemed highly likely to be fraudulent, the server immediately initiates a recording process. This stores the call content as evidence and generates a warning message for the user. This message generation uses synthesized speech technology (e.g., Amazon Polly). The warning is sent directly to both the caller and the recipient in real time during the call.

[0290] After the call, the server notifies the user of the details of the fraud assessment and the recorded data, prompting them to review it. The notification is sent via email or application push notification, and the user can review the content on a secure portal site or dedicated application. This information allows the user to review the call content and take steps to prevent becoming a victim of fraud.

[0291] Furthermore, the server automatically and regularly updates its database of fraudulent talk patterns and continuously learns to counter new fraudulent tactics using its generating AI model. This ensures that the system always reflects the latest fraud information and maintains a form that can adapt to ever-evolving fraudulent methods.

[0292] For example, if a call comes in from someone impersonating a "son" attempting to commit a wire transfer scam, the server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. At this stage, recording begins and a warning is sent to the user.

[0293] Example of a prompt:

[0294] "Is this call potentially a scam? The keywords included are 'accident' and 'urgent transfer'."

[0295] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0296] Step 1:

[0297] When a call is received, the terminal activates its built-in microphone and collects the call audio in real time. This audio data is the input. In other words, the terminal converts the audio signal into a digital format and sends it to the server in the appropriate format. The audio data, as output, is then moved to the next processing stage using real-time transmission technology.

[0298] Step 2:

[0299] The server receives audio data sent from the terminal. Next, a speech recognition component converts this audio data into text. The input data is an audio signal, and the output data is text in string format. Speech recognition technology is used to extract the content of the audio as textual information.

[0300] Step 3:

[0301] The server uses the converted text data to compare it against a database containing registered fraudulent talk patterns. Here, a pattern recognition algorithm is used to detect characteristic phrases of fraud and determine the likelihood of it being a scam. The input data is text, and the output data is a numerical rating of the likelihood of fraud.

[0302] Step 4:

[0303] If the possibility of fraud exceeds the threshold, the server starts the recording process. The input is the result of the fraud judgment. As output, the recorded voice data is stored in the storage system. At the same time, a warning message is generated and notified to the user and the caller using text-to-speech technology. This alerts the recipient.

[0304] Step 5:

[0305] After the call ends, the server notifies the user of the judgment result and the recording data. The input is the recorded data and the judgment result, and the output is the notification information to the user. The notification is sent via email or an application, and the user can check the content on a dedicated portal site or application.

[0306] Step 6:

[0307] The server automatically and regularly updates the fraud speech pattern database and uses the generated AI model to flexibly respond to new fraud methods. The input to this process is the updated fraud information and past data trends, and the output is the updated database and improved model. Thus, the system can always operate with the latest fraud information.

[0308] (Application Example 1)

[0309] Next, Application Example 1 will be described. In the following description, the data processing device 12 is referred to as the "server", and the smart glasses 214 are referred to as the "terminal".

[0310] With the advancement of communication technology, fraudulent activities have diversified, and scams exploiting voice calls are on the rise. Such fraudulent activities are a serious problem, making it difficult for users to use voice calls with peace of mind. In particular, vulnerable groups such as the elderly are at risk of suffering significant losses due to the lack of adequate countermeasures against fraud. In this situation, there is a need for an efficient system that prevents fraudulent activities and provides users with a sense of security.

[0311] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0312] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for acquiring the audio of the incoming communication and processing the audio data in real time, and means for converting the acquired audio into text information and comparing it with fraudulent talk patterns. This enables the automatic detection and warning of potential fraudulent activity, allowing users to use voice calls with peace of mind.

[0313] "Incoming caller" refers to the identification information of the party that initiated the call or message in a communication method.

[0314] "Audio data" refers to information that represents sound in digital or analog format.

[0315] "Textual information" refers to text data obtained by converting data such as audio into a string of characters.

[0316] "Fraudulent behavior talk patterns" are a collection of characteristic voices and phrases extracted from past cases of fraudulent activity.

[0317] "Recording" is the process of saving certain data or information to prepare for future use.

[0318] A "warning" is a message issued to draw attention to risks or potential dangers.

[0319] A "warning audio message" is an audio notification used to warn or caution the sender or recipient when fraudulent activity is suspected.

[0320] A "user" is an entity that operates or utilizes a system, product, or service.

[0321] A "dedicated information terminal" is an electronic device designed specifically for a particular purpose or function.

[0322] An "application program" is software developed to achieve a specific function or purpose.

[0323] The system that realizes this invention mainly consists of three entities: a server, a terminal, and a user. The server is responsible for monitoring calls from unregistered incoming sources and processing those calls in real time. The terminal acquires the audio data of the call and sends it as a stream to the server. The server receives this and converts the audio into text data using the Google Cloud Speech-to-Text API. Based on this text data, it is compared against fraudulent talk patterns, and an AI model using TensorFlow or PyTorch analyzes and determines the possibility of fraud.

[0324] If the server detects a potential for fraud exceeding a threshold, the call is recorded, and a warning message is displayed on the user's device. Additionally, a cautionary voice message is generated and sent to the caller. This process allows the user to continue the call with confidence and later review the call content.

[0325] As a concrete example, consider a scenario where a user receives a phone call saying, "You've won a prize, please provide your bank account information." This system detects and analyzes phrases like "prize" and "bank account information," and if it determines that there is a high probability of fraudulent activity, it immediately starts recording and warns the user that "this may be a scam." An example of a prompt message would be, "Analyze the content of the call received by the user, analyze whether it contains the key phrases 'prize' and 'bank account information,' and issue a warning if it is suspicious."

[0326] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0327] Step 1:

[0328] The terminal detects an incoming call from an unregistered source and acquires audio data as soon as the call begins. This audio data serves as the input and is streamed to the server in real time.

[0329] Step 2:

[0330] The server converts the received audio data into text information using the Google Cloud Speech-to-Text API. This process outputs the audio data as text data.

[0331] Step 3:

[0332] The server compares the generated text data with a database of fraudulent talk patterns. A generative AI model using TensorFlow or PyTorch performs a probabilistic determination by comparing it with a database based on past cases. This process outputs a numerical result indicating the likelihood of fraudulent activity.

[0333] Step 4:

[0334] The server determines a threshold based on the quantified result, and if the likelihood of fraudulent activity exceeds the threshold, it starts recording the call. The input here is the judgment result, and the output is the start of the recording process.

[0335] Step 5:

[0336] A warning message based on the judgment result is displayed on the user's terminal. Warning information from the server is input and output to the user in the form of a notification.

[0337] Step 6:

[0338] The server generates a warning voice message and sends it to the caller. This process outputs a generated voice file, which is then played back by the caller to deliver the warning.

[0339] Step 7:

[0340] After the call ends, the server sends the judgment result and recorded data to the user's dedicated information terminal or application program for user review. In this step, the recorded data is the input, and it is output by being displayed on the user's review interface.

[0341] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0342] This invention combines an emotion engine with a fraud prevention system, providing a function to detect the user's emotions in real time during a call and more accurately determine the likelihood of fraud. This system monitors incoming calls from unregistered callers and analyzes the call content and the user's emotions to improve its ability to respond to fraud.

[0343] When a call comes in to a user's device, if it's from an unregistered caller, the server activates the emotion engine simultaneously with audio collection. Using speech recognition technology, it converts the audio into text and analyzes the user's emotional state from their voice patterns while comparing it against fraudulent talk patterns.

[0344] The device captures audio during a call with high accuracy and sends the necessary data to the server for the emotion engine. Factors such as tone, speed, and emphasis are considered.

[0345] The server uses an emotion engine to detect the user's emotions. The detected emotion data is fed back into the fraud possibility assessment and contributes to an overall decision. This result influences the fraud threshold determination and, if necessary, initiates recording or issues a warning.

[0346] If a user is emotionally distressed, the server will detect this and adjust the content and timing of the warning notification to help the user understand. For example, an additional warning in a calmer tone may be issued.

[0347] After the call ends, the server notifies the user of the detected emotion data and analysis results. This allows the user to understand their own emotional responses and use that information to improve future interactions.

[0348] Emotional data is stored as part of an ongoing learning process, contributing to the updating of the fraudulent talk pattern database and improving the accuracy of the AI ​​model. Using this learning process, the system continues to evolve and adapt to new fraudulent tactics and user emotional patterns.

[0349] For example, if someone claiming to be from a "financial institution" requests "account information verification," and the server detects that the user's voice indicates distress, the server will intensify the warning and provide an interaction prompting the user to acknowledge their emotions and exercise caution.

[0350] The following describes the processing flow.

[0351] Step 1:

[0352] The server monitors incoming calls to the user's terminal and detects calls from unregistered callers. This detection triggers the activation of voice data processing and the emotion engine.

[0353] Step 2:

[0354] As soon as a call starts, the device captures voice data with high precision and sends it to the server in stream format, including the voice parameters necessary for the emotion engine.

[0355] Step 3:

[0356] The server uses speech recognition technology to convert the audio data into text and compares it against a database of fraudulent talk patterns. In parallel, an emotion engine analyzes the tone, speed, emphasis, and other factors of the voice.

[0357] Step 4:

[0358] The server receives analysis results from the emotion engine and provides feedback to the fraud detection score if the user's emotions indicate anxiety or distress.

[0359] Step 5:

[0360] If the fraud probability score exceeds a threshold, the server will start recording and notify the user with a warning message tailored to their emotional state.

[0361] Step 6:

[0362] If a user shows emotional changes during a call, the server detects this in real time and dynamically adjusts the content and intensity of warning notifications to support user understanding.

[0363] Step 7:

[0364] After the call ends, the server notifies the user of the results, audio data, and sentiment analysis results. The user can then view this information through a dedicated portal or application.

[0365] Step 8:

[0366] The server updates the database using sentiment data and call history, and the AI ​​model continues to learn from the new data. This process allows the system to evolve and improve the accuracy of future fraud detections.

[0367] (Example 2)

[0368] Next, we will describe Example 2. 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".

[0369] Traditional fraud prevention systems relied on standard voice recognition and simple pattern matching for communications from unregistered sources, making it difficult to accurately determine the likelihood of fraud. Furthermore, they could not take into account the user's emotional state, resulting in inappropriate timing and content of warnings regarding fraudulent methods. Additionally, they struggled to respond quickly to the emergence of new fraud techniques.

[0370] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0371] In this invention, the server includes means for capturing transmitted audio and processing the audio information in real time, means for converting the captured audio into text and comparing it with fraud patterns, and means for analyzing the characteristics of the audio and determining the emotional state. This makes it possible to determine the possibility of fraud with high accuracy while taking into account the user's emotional state and to provide appropriate warnings and advice in real time. Furthermore, by regularly updating the information source for fraudulent talk patterns and improving accuracy through a learning device, it is possible to respond quickly to new fraudulent methods.

[0372] "Means for detecting communications" refers to a function that checks for and detects the presence of calls and messages from unregistered callers.

[0373] "Means for processing voice information" refers to a function that receives voice during a call and converts and manages it as digital data in real time.

[0374] "Means of converting to text" refers to a function that performs the process of converting audio data into text information using speech recognition technology.

[0375] "Methods for matching with fraud patterns" refers to a function that compares transcribed audio information with known fraudulent talk patterns and analyzes the degree of similarity.

[0376] "A means of analyzing voice characteristics and determining emotional state" refers to a function that analyzes features such as tone, speed, and emphasis of voice to identify the user's emotional response.

[0377] "Means of providing warnings and advice" refers to a function that alerts users and provides specific countermeasures when a potential scam is detected.

[0378] "Methods for regularly updating information sources" refers to the process of keeping fraud databases up-to-date and incorporating new fraud patterns.

[0379] A "learning device" is a technology necessary to train an AI model based on new information and continuously improve the system's discrimination accuracy.

[0380] This invention is a system aimed at preventing fraud, providing a function to analyze the user's emotions in real time during communication and determine the possibility of fraud. To achieve this, a system configuration in which a server and terminals are linked is necessary.

[0381] The server receives communications from the user's terminal. Specific software used includes speech recognition technology and generative AI models. The server transcribes the received audio data into text in real time and compares it against a fraud pattern database. During this process, features extracted from the audio data are used to activate an emotion engine and analyze the user's emotional state.

[0382] The terminal is responsible for capturing audio during calls with high precision. Hardware components include a microphone and an audio processing unit. The terminal converts the audio signal into digital data and sends it to the server. This allows the server to perform more accurate analysis.

[0383] As a concrete example, consider a scenario where someone claiming to be from a "financial institution" requests "confirmation of account information" from an unknown caller. In this situation, where the user is inevitably suspicious of fraud, the system detects the user's emotional distress and immediately intensifies the warning. The user is given calm voice advice such as, "Stay calm and do not carelessly provide personal information." This process is achieved by instructing the system to "detect keywords suggesting potential fraud through speech recognition and also perform emotional analysis of the user's emotional state."

[0384] In this system, emotional data is accumulated through a continuous learning process, and the fraudulent talk pattern database is also updated regularly. This allows the system to continuously adapt to new fraudulent tactics and changing user emotional patterns. By receiving this feedback, users can better understand their future unconscious reactions and take countermeasures.

[0385] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0386] Step 1:

[0387] The terminal detects incoming calls. Inputs include the caller's number and the call start signal. If the call is from an unregistered caller, the terminal immediately begins capturing audio data. The audio is captured by the microphone, noise is removed to produce clear audio data, and it is converted to a digital format for output.

[0388] Step 2:

[0389] The terminal transmits the captured audio data to the server in real time. During this process, acoustic characteristics such as tone, speed, and emphasis are input and sent directly to the server as digital data.

[0390] Step 3:

[0391] The server performs speech recognition based on the received audio data. Using a generative AI model, it converts speech to text, outputting text data from the audio data as input. During this process, advanced algorithms are employed to perform precise analysis of the audio data.

[0392] Step 4:

[0393] The server compares the transcribed text against a database of fraud patterns. It compares the text data received as input against known fraud patterns and outputs a degree of similarity based on the results. Based on the matching results, it makes an initial determination of the likelihood of fraud.

[0394] Step 5:

[0395] The server activates an emotion engine to analyze the user's emotions from the voice data. It takes voice characteristic data as input and outputs an emotional state (e.g., calm, agitated). The emotion analysis results are fed back into the fraud detection process.

[0396] Step 6:

[0397] The server comprehensively assesses the likelihood of fraud, issues a warning if necessary, and begins recording. In this step, it uses whether the assessed fraud likelihood score exceeds a threshold as a criterion, and adjusts the content and timing of the warning notification before sending it to the user. The output includes the warning content and stored recording data.

[0398] Step 7:

[0399] The server provides emotional feedback to the user after the call ends. Using the analyzed emotional data and fraud detection results as input, it outputs analysis results as appropriate and notifies the user via their terminal or a dedicated digital platform. Based on this, the user can consider future countermeasures.

[0400] Step 8:

[0401] The server feeds back the collected data to the learning device, updating the generated AI model. This process involves taking new fraud patterns and emotional patterns as input and outputting improvements in AI accuracy and system evolution.

[0402] (Application Example 2)

[0403] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0404] In recent years, telephone fraud tactics have become increasingly sophisticated, with a particular emphasis on appealing to emotions. Traditional fraud prevention systems rely solely on matching voice patterns and keywords, failing to consider the user's emotional state, which can lead to false positives and user confusion. Furthermore, users rarely have the opportunity to recognize their own emotional responses during a call, making it difficult for them to take preventative measures.

[0405] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0406] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and processing the audio data in real time, and means for analyzing the user's emotions from their voice patterns and determining the likelihood of fraud based on the emotional data. This makes it possible to more accurately assess the risk of fraud while recognizing the user's emotional response and to take warnings and risk avoidance actions.

[0407] "Unregistered caller" refers to an incoming call from a phone number that is not pre-registered on the user's device or system.

[0408] "Audio data processing" refers to a series of operations that analyze or transform audio acquired during a call for a specific purpose.

[0409] "Converting speech to text" is the process of automatically replacing spoken information generated during a phone call with written text.

[0410] A "fraudulent talk pattern" refers to the characteristics of audio data that include conversational content or specific phrases intended for fraudulent purposes.

[0411] "Emotional data" refers to information that quantifies or represents the emotional tendencies and reactions that a user exhibits during a call.

[0412] "Notifying a warning" refers to providing users with information to inform them of the risks and encourage them to take safety measures when fraud is suspected.

[0413] A "learning model" is an algorithm or system that uses machine learning to recognize and predict specific data patterns.

[0414] "Emotional feedback" is an information provision process that informs users about the content and impact of their emotional responses during a call.

[0415] This invention is a system that uses the user's communication device to analyze the possibility of fraud in real time on the server side and issue warnings based on sentiment data. To implement this system, the user's terminal first detects an incoming call from an unregistered source. When a voice call begins, the terminal captures the audio during the call with high accuracy and transmits it to the server in real time.

[0416] The server uses speech recognition technologies such as the Google Cloud Speech-to-Text API to convert received audio into text data. This text data is compared against a database of fraudulent talk patterns and analyzed to determine the likelihood of fraud. Simultaneously, the server uses the Python transformers library to analyze the user's emotional state from the transcribed conversation. The emotional data obtained in this process is fed back into the judgment process.

[0417] If the server determines that a user is facing a risk of fraud, it sends a warning notification to the user's device via the Flutter framework, and the audio is recorded. This allows the user to understand the risks during the call and take appropriate action. After the call ends, the server also provides the user with emotional feedback and encourages self-analysis of the past call through a dedicated application.

[0418] For example, if a user is approached with a fraudulent financial transaction offer, and the audio matches a scammer's conversation pattern, and the system detects signs of distress in the user, it will immediately issue a warning and begin recording. This allows the user to gain knowledge about the fraudulent activity and strengthen their preparations for the next steps.

[0419] An example of a prompt to a generative AI model is: "Assess the likelihood of fraud based on the following sentiment data and text in response to the caller's call, and output the result. Sentiment data: agitated, frightened. Text: 'Please provide your account information.'" By using this prompt, the AI ​​model can gain a deeper understanding of the caller's intent and effectively assess the risk of fraud.

[0420] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0421] Step 1:

[0422] The terminal detects incoming calls from unregistered callers. At this time, it receives the caller's number information as input, and if it does not exist in the internal registration database, it determines the caller is unknown and activates the system.

[0423] Step 2:

[0424] The terminal captures audio during a call and sends it to the server with high accuracy. It receives audio data as input, encodes it, and sends it to the server in real time. Here, noise cancellation technology is used to maintain clear audio data.

[0425] Step 3:

[0426] The server uses the Google Cloud Speech-to-Text API to convert audio data into text. The server receives audio data as input and applies speech recognition technology to generate text output. This output text forms the basis for the next analysis step.

[0427] Step 4:

[0428] The server uses text data to match it against scam talk patterns. Here, it takes text data as input and compares it with a scam database to perform data calculations that identify phrases that match predefined scam patterns.

[0429] Step 5:

[0430] The server uses the transformers library to analyze the user's emotional state from the text. This process quantifies the emotional data extracted from the input text and classifies it into categories such as positive, negative, and neutral. The calculated emotional data is then used in the next decision step.

[0431] Step 6:

[0432] The server determines the risk of fraud based on the degree of match with fraud patterns and the user's sentiment data. It receives sentiment data and fraud match scores as input, compares them to a threshold, and outputs whether a warning is necessary. Based on this result, it decides whether a warning is sent to the user.

[0433] Step 7:

[0434] If the server determines that the level has exceeded the danger threshold, it sends a warning notification to the terminal and starts recording. At this point, it uses the final risk assessment result as input to generate a warning message, displays the warning on the terminal, and starts the process of saving the audio data.

[0435] Step 8:

[0436] After the call ends, the server generates emotional feedback and notifies the user. The emotional data collected during the call is aggregated and converted into a feedback format that is easy for the user to understand. This result is then communicated to the user and used to improve future fraud prevention efforts.

[0437] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0438] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0439] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0440] [Third Embodiment]

[0441] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0442] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0443] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0444] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0445] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0446] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0447] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0448] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0449] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0450] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0451] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0452] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0453] This invention is a fraud prevention system that monitors incoming calls from unregistered callers in real time and determines the possibility of fraud, thereby preventing damage. The operation of the entire system is described below.

[0454] The server constantly monitors incoming calls to the user's device. If a call comes in from a number other than those registered by the user, it automatically initiates an AI-based analysis process.

[0455] The terminal collects audio data as soon as a call is received and streams it to the server. The transmitted data is converted to text using speech recognition technology, making the call content available for analysis.

[0456] The server converts the audio to text and then compares the text against a database containing fraudulent talk patterns. This process applies a pattern recognition algorithm based on past fraud cases, quantifying the likelihood of fraud.

[0457] When the likelihood of fraud exceeds a certain threshold, the server initiates a recording process. This allows the call to be stored as evidence. Simultaneously, a warning message is generated in audio form and notified to both parties involved in the call, alerting the recipient and providing a psychological deterrent to the caller.

[0458] After the call, the server notifies the user of the details of the fraud detection and provides the recording data. This feature is designed to allow users to review the call content later, providing them with peace of mind.

[0459] The servers regularly update their fraud database and use AI learning models to flexibly adapt to new fraudulent methods. This automated update process ensures the system is always up-to-date, protecting users from fraud over the long term.

[0460] As a concrete example, consider a scenario where a caller attempts to commit a wire fraud by impersonating the user's "son." The server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. As a result, recording begins immediately, and the user is notified with a warning. This process effectively reduces the risk of the user being deceived by a fraudulent call.

[0461] The following describes the processing flow.

[0462] Step 1:

[0463] The server monitors all incoming calls to the user's device and detects incoming calls from unregistered callers. This detection triggers the start of AI processing.

[0464] Step 2:

[0465] As soon as a call begins, the terminal collects voice data in real time and streams it to the server. During this process, the voice data is compressed to optimize bandwidth usage.

[0466] Step 3:

[0467] The server converts the received audio data into text using speech recognition technology. It also performs noise reduction processing to improve audio clarity.

[0468] Step 4:

[0469] The server compares the converted text against a database of scam talk patterns. Using natural language processing algorithms, it compares extracted key phrases and performs pattern recognition.

[0470] Step 5:

[0471] The server scores the likelihood of fraud. If the calculated score exceeds a threshold, it determines that there is a high probability of fraud and proceeds to the next step.

[0472] Step 6:

[0473] The server starts recording as soon as it determines that a call is highly likely to be a scam. This recording is saved as evidence and remains accessible to the user even after the call ends.

[0474] Step 7:

[0475] The server uses speech synthesis technology to generate a warning message and notifies both parties in the call in real time. This draws the listener's attention and puts pressure on the caller.

[0476] Step 8:

[0477] After the call ends, the server notifies the user of the fraud detection result and guides them to view the recorded data on a dedicated portal or application.

[0478] Step 9:

[0479] The server updates the fraudulent talk pattern database based on call data and judgment results. The AI ​​model continuously learns using newly collected data to improve its accuracy.

[0480] (Example 1)

[0481] Next, we will describe Example 1. 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."

[0482] In modern times, telephone fraud is becoming increasingly sophisticated year by year. This raises the risk of ordinary consumers unintentionally becoming victims of fraud, posing a significant security challenge, especially for the elderly and individuals with low IT literacy. Traditional methods such as manual registration restrictions and simple suspicious number blocking are insufficient; therefore, a comprehensive fraud prevention system employing new methods is needed.

[0483] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0484] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and streaming the audio data in real time using a data processing device, and means for converting the captured audio into text using speech recognition technology and comparing it with fraudulent talk patterns. This enables automatic detection of fraudulent activity and immediate response to it.

[0485] "Unregistered caller ID" refers to phone numbers or contact information that the user has not previously registered in the system.

[0486] "Means for detecting incoming calls" refers to a device or method that monitors telephone communications and identifies calls from unregistered callers.

[0487] "Means for capturing audio" refers to a device or mechanism for capturing audio signals generated during a telephone call and inputting them into a system.

[0488] A "data processing device" refers to a hardware or software component used for processing, converting, and analyzing audio signals.

[0489] "Means for streaming audio data" refers to a technology or method for transmitting acquired audio data in real time to other devices or systems.

[0490] "Speech recognition technology" refers to the technology that receives speech signals and converts them into text format that can be understood as human language.

[0491] "Fraudulent talk patterns" refer to typical conversational content and phrase patterns from past fraud cases, and are used as criteria for diagnosing the possibility of fraud.

[0492] A "pattern recognition algorithm" refers to a program or method that identifies certain regularities or features in data and performs classification or judgment based on specific conditions.

[0493] A "generative AI model" refers to an artificial intelligence model that learns new data patterns and uses them to generate data and make inferences.

[0494] By using the following configuration and technology in the implementation of this fraud prevention system, it is possible to smoothly and efficiently detect potential fraud and protect users.

[0495] First, the device receives an incoming call through the user's phone function. Upon receiving the call, the device's microphone is activated, and audio data from the call is collected in real time. This audio data is immediately streamed to the server via the communication network. Real-time data processing technologies such as WebRTC are used for streaming.

[0496] The server uses the received audio data to convert it to text using speech recognition software (e.g., Google Cloud Speech-to-Text). The transcribed data is then compared against a database containing recorded fraudulent talk patterns. Here, a pattern recognition algorithm using a natural language processing library (e.g., NLTK or Spacy) is applied to quantify the likelihood of fraud.

[0497] If a call is deemed highly likely to be fraudulent, the server immediately initiates a recording process. This stores the call content as evidence and generates a warning message for the user. This message generation uses synthesized speech technology (e.g., Amazon Polly). The warning is sent directly to both the caller and the recipient in real time during the call.

[0498] After the call, the server notifies the user of the details of the fraud assessment and the recorded data, prompting them to review it. The notification is sent via email or application push notification, and the user can review the content on a secure portal site or dedicated application. This information allows the user to review the call content and take steps to prevent becoming a victim of fraud.

[0499] Furthermore, the server automatically and regularly updates its database of fraudulent talk patterns and continuously learns to counter new fraudulent tactics using its generating AI model. This ensures that the system always reflects the latest fraud information and maintains a form that can adapt to ever-evolving fraudulent methods.

[0500] For example, if a call comes in from someone impersonating a "son" attempting to commit a wire transfer scam, the server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. At this stage, recording begins and a warning is sent to the user.

[0501] Example of a prompt:

[0502] "Is this call potentially a scam? The keywords included are 'accident' and 'urgent transfer'."

[0503] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0504] Step 1:

[0505] When a call is received, the terminal activates its built-in microphone and collects the call audio in real time. This audio data is the input. In other words, the terminal converts the audio signal into a digital format and sends it to the server in the appropriate format. The audio data, as output, is then moved to the next processing stage using real-time transmission technology.

[0506] Step 2:

[0507] The server receives audio data sent from the terminal. Next, a speech recognition component converts this audio data into text. The input data is an audio signal, and the output data is text in string format. Speech recognition technology is used to extract the content of the audio as textual information.

[0508] Step 3:

[0509] The server uses the converted text data to compare it against a database containing registered fraudulent talk patterns. Here, a pattern recognition algorithm is used to detect characteristic phrases of fraud and determine the likelihood of it being a scam. The input data is text, and the output data is a numerical rating of the likelihood of fraud.

[0510] Step 4:

[0511] If the likelihood of fraud exceeds a certain threshold, the server initiates the recording process. The input is the result of the fraud detection. As output, the recorded audio data is saved to the storage system. Simultaneously, a warning message is generated and sent to the user and caller using synthesized speech technology. This serves to alert the recipient.

[0512] Step 5:

[0513] After the call ends, the server notifies the user of the judgment result and the recorded data. The input is the recorded data and the judgment result, and the output is the notification information to the user. The notification is sent via email or application, and the user can check the content on a dedicated portal site or application.

[0514] Step 6:

[0515] The server automatically and regularly updates its database of fraudulent talk patterns and uses a generating AI model to flexibly adapt to new fraudulent methods. The input to this process is updated fraud information and historical data trends, while the output is an updated database and an improved model. In this way, the system can always operate with the most up-to-date fraud information.

[0516] (Application Example 1)

[0517] Next, we will explain Application Example 1. In the following explanation, 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."

[0518] With the advancement of communication technology, fraudulent activities have diversified, and scams exploiting voice calls are on the rise. Such fraudulent activities are a serious problem, making it difficult for users to use voice calls with peace of mind. In particular, vulnerable groups such as the elderly are at risk of suffering significant losses due to the lack of adequate countermeasures against fraud. In this situation, there is a need for an efficient system that prevents fraudulent activities and provides users with a sense of security.

[0519] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0520] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for acquiring the audio of the incoming communication and processing the audio data in real time, and means for converting the acquired audio into text information and comparing it with fraudulent talk patterns. This enables the automatic detection and warning of potential fraudulent activity, allowing users to use voice calls with peace of mind.

[0521] "Incoming caller" refers to the identification information of the party that initiated the call or message in a communication method.

[0522] "Audio data" refers to information that represents sound in digital or analog format.

[0523] "Textual information" refers to text data obtained by converting data such as audio into a string of characters.

[0524] "Fraudulent behavior talk patterns" are a collection of characteristic voices and phrases extracted from past cases of fraudulent activity.

[0525] "Recording" is the process of saving certain data or information to prepare for future use.

[0526] A "warning" is a message issued to draw attention to risks or potential dangers.

[0527] A "warning audio message" is an audio notification used to warn or caution the sender or recipient when fraudulent activity is suspected.

[0528] A "user" is an entity that operates or utilizes a system, product, or service.

[0529] A "dedicated information terminal" is an electronic device designed specifically for a particular purpose or function.

[0530] An "application program" is software developed to achieve a specific function or purpose.

[0531] The system that realizes this invention mainly consists of three entities: a server, a terminal, and a user. The server is responsible for monitoring calls from unregistered incoming sources and processing those calls in real time. The terminal acquires the audio data of the call and sends it as a stream to the server. The server receives this and converts the audio into text data using the Google Cloud Speech-to-Text API. Based on this text data, it is compared against fraudulent talk patterns, and an AI model using TensorFlow or PyTorch analyzes and determines the possibility of fraud.

[0532] If the server detects a potential for fraud exceeding a threshold, the call is recorded, and a warning message is displayed on the user's device. Additionally, a cautionary voice message is generated and sent to the caller. This process allows the user to continue the call with confidence and later review the call content.

[0533] As a concrete example, consider a scenario where a user receives a phone call saying, "You've won a prize, please provide your bank account information." This system detects and analyzes phrases like "prize" and "bank account information," and if it determines that there is a high probability of fraudulent activity, it immediately starts recording and warns the user that "this may be a scam." An example of a prompt message would be, "Analyze the content of the call received by the user, analyze whether it contains the key phrases 'prize' and 'bank account information,' and issue a warning if it is suspicious."

[0534] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0535] Step 1:

[0536] The terminal detects an incoming call from an unregistered source and acquires audio data as soon as the call begins. This audio data serves as the input and is streamed to the server in real time.

[0537] Step 2:

[0538] The server converts the received audio data into text information using the Google Cloud Speech-to-Text API. This process outputs the audio data as text data.

[0539] Step 3:

[0540] The server compares the generated text data with a database of fraudulent talk patterns. A generative AI model using TensorFlow or PyTorch performs a probabilistic determination by comparing it with a database based on past cases. This process outputs a numerical result indicating the likelihood of fraudulent activity.

[0541] Step 4:

[0542] The server determines a threshold based on the quantified result, and if the likelihood of fraudulent activity exceeds the threshold, it starts recording the call. The input here is the judgment result, and the output is the start of the recording process.

[0543] Step 5:

[0544] A warning message based on the judgment result is displayed on the user's terminal. Warning information from the server is input and output to the user in the form of a notification.

[0545] Step 6:

[0546] The server generates a warning voice message and sends it to the caller. This process outputs a generated voice file, which is then played back by the caller to deliver the warning.

[0547] Step 7:

[0548] After the call ends, the server sends the judgment result and recorded data to the user's dedicated information terminal or application program for user review. In this step, the recorded data is the input, and it is output by being displayed on the user's review interface.

[0549] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0550] This invention combines an emotion engine with a fraud prevention system, providing a function to detect the user's emotions in real time during a call and more accurately determine the likelihood of fraud. This system monitors incoming calls from unregistered callers and analyzes the call content and the user's emotions to improve its ability to respond to fraud.

[0551] When a call comes in to a user's device, if it's from an unregistered caller, the server activates the emotion engine simultaneously with audio collection. Using speech recognition technology, it converts the audio into text and analyzes the user's emotional state from their voice patterns while comparing it against fraudulent talk patterns.

[0552] The device captures audio during a call with high accuracy and sends the necessary data to the server for the emotion engine. Factors such as tone, speed, and emphasis are considered.

[0553] The server uses an emotion engine to detect the user's emotions. The detected emotion data is fed back into the fraud possibility assessment and contributes to an overall decision. This result influences the fraud threshold determination and, if necessary, initiates recording or issues a warning.

[0554] If a user is emotionally distressed, the server will detect this and adjust the content and timing of the warning notification to help the user understand. For example, an additional warning in a calmer tone may be issued.

[0555] After the call ends, the server notifies the user of the detected emotion data and analysis results. This allows the user to understand their own emotional responses and use that information to improve future interactions.

[0556] Emotional data is stored as part of an ongoing learning process, contributing to the updating of the fraudulent talk pattern database and improving the accuracy of the AI ​​model. Using this learning process, the system continues to evolve and adapt to new fraudulent tactics and user emotional patterns.

[0557] For example, if someone claiming to be from a "financial institution" requests "account information verification," and the server detects that the user's voice indicates distress, the server will intensify the warning and provide an interaction prompting the user to acknowledge their emotions and exercise caution.

[0558] The following describes the processing flow.

[0559] Step 1:

[0560] The server monitors incoming calls to the user's terminal and detects calls from unregistered callers. This detection triggers the activation of voice data processing and the emotion engine.

[0561] Step 2:

[0562] As soon as a call starts, the device captures voice data with high precision and sends it to the server in stream format, including the voice parameters necessary for the emotion engine.

[0563] Step 3:

[0564] The server uses speech recognition technology to convert the audio data into text and compares it against a database of fraudulent talk patterns. In parallel, an emotion engine analyzes the tone, speed, emphasis, and other factors of the voice.

[0565] Step 4:

[0566] The server receives analysis results from the emotion engine and provides feedback to the fraud detection score if the user's emotions indicate anxiety or distress.

[0567] Step 5:

[0568] If the fraud probability score exceeds a threshold, the server will start recording and notify the user with a warning message tailored to their emotional state.

[0569] Step 6:

[0570] If a user shows emotional changes during a call, the server detects this in real time and dynamically adjusts the content and intensity of warning notifications to support user understanding.

[0571] Step 7:

[0572] After the call ends, the server notifies the user of the results, audio data, and sentiment analysis results. The user can then view this information through a dedicated portal or application.

[0573] Step 8:

[0574] The server updates the database using sentiment data and call history, and the AI ​​model continues to learn from the new data. This process allows the system to evolve and improve the accuracy of future fraud detections.

[0575] (Example 2)

[0576] Next, we will describe Example 2. 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."

[0577] Traditional fraud prevention systems relied on standard voice recognition and simple pattern matching for communications from unregistered sources, making it difficult to accurately determine the likelihood of fraud. Furthermore, they could not take into account the user's emotional state, resulting in inappropriate timing and content of warnings regarding fraudulent methods. Additionally, they struggled to respond quickly to the emergence of new fraud techniques.

[0578] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0579] In this invention, the server includes means for capturing transmitted audio and processing the audio information in real time, means for converting the captured audio into text and comparing it with fraud patterns, and means for analyzing the characteristics of the audio and determining the emotional state. This makes it possible to determine the possibility of fraud with high accuracy while taking into account the user's emotional state and to provide appropriate warnings and advice in real time. Furthermore, by regularly updating the information source for fraudulent talk patterns and improving accuracy through a learning device, it is possible to respond quickly to new fraudulent methods.

[0580] "Means for detecting communications" refers to a function that checks for and detects the presence of calls and messages from unregistered callers.

[0581] "Means for processing voice information" refers to a function that receives voice during a call and converts and manages it as digital data in real time.

[0582] "Means of converting to text" refers to a function that performs the process of converting audio data into text information using speech recognition technology.

[0583] "Methods for matching with fraud patterns" refers to a function that compares transcribed audio information with known fraudulent talk patterns and analyzes the degree of similarity.

[0584] "A means of analyzing voice characteristics and determining emotional state" refers to a function that analyzes features such as tone, speed, and emphasis of voice to identify the user's emotional response.

[0585] "Means of providing warnings and advice" refers to a function that alerts users and provides specific countermeasures when a potential scam is detected.

[0586] "Methods for regularly updating information sources" refers to the process of keeping fraud databases up-to-date and incorporating new fraud patterns.

[0587] A "learning device" is a technology necessary to train an AI model based on new information and continuously improve the system's discrimination accuracy.

[0588] This invention is a system aimed at preventing fraud, providing a function to analyze the user's emotions in real time during communication and determine the possibility of fraud. To achieve this, a system configuration in which a server and terminals are linked is necessary.

[0589] The server receives communications from the user's terminal. Specific software used includes speech recognition technology and generative AI models. The server transcribes the received audio data into text in real time and compares it against a fraud pattern database. During this process, features extracted from the audio data are used to activate an emotion engine and analyze the user's emotional state.

[0590] The terminal is responsible for capturing audio during calls with high precision. Hardware components include a microphone and an audio processing unit. The terminal converts the audio signal into digital data and sends it to the server. This allows the server to perform more accurate analysis.

[0591] As a concrete example, consider a scenario where someone claiming to be from a "financial institution" requests "confirmation of account information" from an unknown caller. In this situation, where the user is inevitably suspicious of fraud, the system detects the user's emotional distress and immediately intensifies the warning. The user is given calm voice advice such as, "Stay calm and do not carelessly provide personal information." This process is achieved by instructing the system to "detect keywords suggesting potential fraud through speech recognition and also perform emotional analysis of the user's emotional state."

[0592] In this system, emotional data is accumulated through a continuous learning process, and the fraudulent talk pattern database is also updated regularly. This allows the system to continuously adapt to new fraudulent tactics and changing user emotional patterns. By receiving this feedback, users can better understand their future unconscious reactions and take countermeasures.

[0593] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0594] Step 1:

[0595] The terminal detects incoming calls. Inputs include the caller's number and the call start signal. If the call is from an unregistered caller, the terminal immediately begins capturing audio data. The audio is captured by the microphone, noise is removed to produce clear audio data, and it is converted to a digital format for output.

[0596] Step 2:

[0597] The terminal transmits the captured audio data to the server in real time. During this process, acoustic characteristics such as tone, speed, and emphasis are input and sent directly to the server as digital data.

[0598] Step 3:

[0599] The server performs speech recognition based on the received audio data. Using a generative AI model, it converts speech to text, outputting text data from the audio data as input. During this process, advanced algorithms are employed to perform precise analysis of the audio data.

[0600] Step 4:

[0601] The server compares the transcribed text against a database of fraud patterns. It compares the text data received as input against known fraud patterns and outputs a degree of similarity based on the results. Based on the matching results, it makes an initial determination of the likelihood of fraud.

[0602] Step 5:

[0603] The server activates an emotion engine to analyze the user's emotions from the voice data. It takes voice characteristic data as input and outputs an emotional state (e.g., calm, agitated). The emotion analysis results are fed back into the fraud detection process.

[0604] Step 6:

[0605] The server comprehensively assesses the likelihood of fraud, issues a warning if necessary, and begins recording. In this step, it uses whether the assessed fraud likelihood score exceeds a threshold as a criterion, and adjusts the content and timing of the warning notification before sending it to the user. The output includes the warning content and stored recording data.

[0606] Step 7:

[0607] The server provides emotional feedback to the user after the call ends. Using the analyzed emotional data and fraud detection results as input, it outputs analysis results as appropriate and notifies the user via their terminal or a dedicated digital platform. Based on this, the user can consider future countermeasures.

[0608] Step 8:

[0609] The server feeds back the collected data to the learning device, updating the generated AI model. This process involves taking new fraud patterns and emotional patterns as input and outputting improvements in AI accuracy and system evolution.

[0610] (Application Example 2)

[0611] Next, we will explain application example 2. In the following explanation, 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."

[0612] In recent years, telephone fraud tactics have become increasingly sophisticated, with a particular emphasis on appealing to emotions. Traditional fraud prevention systems rely solely on matching voice patterns and keywords, failing to consider the user's emotional state, which can lead to false positives and user confusion. Furthermore, users rarely have the opportunity to recognize their own emotional responses during a call, making it difficult for them to take preventative measures.

[0613] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0614] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and processing the audio data in real time, and means for analyzing the user's emotions from their voice patterns and determining the likelihood of fraud based on the emotional data. This makes it possible to more accurately assess the risk of fraud while recognizing the user's emotional response and to take warnings and risk avoidance actions.

[0615] "Unregistered caller" refers to an incoming call from a phone number that is not pre-registered on the user's device or system.

[0616] "Audio data processing" refers to a series of operations that analyze or transform audio acquired during a call for a specific purpose.

[0617] "Converting speech to text" is the process of automatically replacing spoken information generated during a phone call with written text.

[0618] A "fraudulent talk pattern" refers to the characteristics of audio data that include conversational content or specific phrases intended for fraudulent purposes.

[0619] "Emotional data" refers to information that quantifies or represents the emotional tendencies and reactions that a user exhibits during a call.

[0620] "Notifying a warning" refers to providing users with information to inform them of the risks and encourage them to take safety measures when fraud is suspected.

[0621] A "learning model" is an algorithm or system that uses machine learning to recognize and predict specific data patterns.

[0622] "Emotional feedback" is an information provision process that informs users about the content and impact of their emotional responses during a call.

[0623] This invention is a system that uses the user's communication device to analyze the possibility of fraud in real time on the server side and issue warnings based on sentiment data. To implement this system, the user's terminal first detects an incoming call from an unregistered source. When a voice call begins, the terminal captures the audio during the call with high accuracy and transmits it to the server in real time.

[0624] The server uses speech recognition technologies such as the Google Cloud Speech-to-Text API to convert received audio into text data. This text data is compared against a database of fraudulent talk patterns and analyzed to determine the likelihood of fraud. Simultaneously, the server uses the Python transformers library to analyze the user's emotional state from the transcribed conversation. The emotional data obtained in this process is fed back into the judgment process.

[0625] If the server determines that a user is facing a risk of fraud, it sends a warning notification to the user's device via the Flutter framework, and the audio is recorded. This allows the user to understand the risks during the call and take appropriate action. After the call ends, the server also provides the user with emotional feedback and encourages self-analysis of the past call through a dedicated application.

[0626] For example, if a user is approached with a fraudulent financial transaction offer, and the audio matches a scammer's conversation pattern, and the system detects signs of distress in the user, it will immediately issue a warning and begin recording. This allows the user to gain knowledge about the fraudulent activity and strengthen their preparations for the next steps.

[0627] An example of a prompt to a generative AI model is: "Assess the likelihood of fraud based on the following sentiment data and text in response to the caller's call, and output the result. Sentiment data: agitated, frightened. Text: 'Please provide your account information.'" By using this prompt, the AI ​​model can gain a deeper understanding of the caller's intent and effectively assess the risk of fraud.

[0628] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0629] Step 1:

[0630] The terminal detects incoming calls from unregistered callers. At this time, it receives the caller's number information as input, and if it does not exist in the internal registration database, it determines the caller is unknown and activates the system.

[0631] Step 2:

[0632] The terminal captures audio during a call and sends it to the server with high accuracy. It receives audio data as input, encodes it, and sends it to the server in real time. Here, noise cancellation technology is used to maintain clear audio data.

[0633] Step 3:

[0634] The server uses the Google Cloud Speech-to-Text API to convert audio data into text. The server receives audio data as input and applies speech recognition technology to generate text output. This output text forms the basis for the next analysis step.

[0635] Step 4:

[0636] The server uses text data to match it against scam talk patterns. Here, it takes text data as input and compares it with a scam database to perform data calculations that identify phrases that match predefined scam patterns.

[0637] Step 5:

[0638] The server uses the transformers library to analyze the user's emotional state from the text. This process quantifies the emotional data extracted from the input text and classifies it into categories such as positive, negative, and neutral. The calculated emotional data is then used in the next decision step.

[0639] Step 6:

[0640] The server determines the risk of fraud based on the degree of match with fraud patterns and the user's sentiment data. It receives sentiment data and fraud match scores as input, compares them to a threshold, and outputs whether a warning is necessary. Based on this result, it decides whether a warning is sent to the user.

[0641] Step 7:

[0642] If the server determines that the level has exceeded the danger threshold, it sends a warning notification to the terminal and starts recording. At this point, it uses the final risk assessment result as input to generate a warning message, displays the warning on the terminal, and starts the process of saving the audio data.

[0643] Step 8:

[0644] After the call ends, the server generates emotional feedback and notifies the user. The emotional data collected during the call is aggregated and converted into a feedback format that is easy for the user to understand. This result is then communicated to the user and used to improve future fraud prevention efforts.

[0645] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0646] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0647] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[0648] [Fourth Embodiment]

[0649] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0650] As shown in Figure 7, the 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.

[0651] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0652] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[0653] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0654] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0655] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0656] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive 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 robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0657] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0658] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0659] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0660] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0661] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0662] This invention is a fraud prevention system that monitors incoming calls from unregistered callers in real time and determines the possibility of fraud, thereby preventing damage. The operation of the entire system is described below.

[0663] The server constantly monitors incoming calls to the user's device. If a call comes in from a number other than those registered by the user, it automatically initiates an AI-based analysis process.

[0664] The terminal collects audio data as soon as a call is received and streams it to the server. The transmitted data is converted to text using speech recognition technology, making the call content available for analysis.

[0665] The server converts the audio to text and then compares the text against a database containing fraudulent talk patterns. This process applies a pattern recognition algorithm based on past fraud cases, quantifying the likelihood of fraud.

[0666] When the likelihood of fraud exceeds a certain threshold, the server initiates a recording process. This allows the call to be stored as evidence. Simultaneously, a warning message is generated in audio form and notified to both parties involved in the call, alerting the recipient and providing a psychological deterrent to the caller.

[0667] After the call, the server notifies the user of the details of the fraud detection and provides the recording data. This feature is designed to allow users to review the call content later, providing them with peace of mind.

[0668] The servers regularly update their fraud database and use AI learning models to flexibly adapt to new fraudulent methods. This automated update process ensures the system is always up-to-date, protecting users from fraud over the long term.

[0669] As a concrete example, consider a scenario where a caller attempts to commit a wire fraud by impersonating the user's "son." The server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. As a result, recording begins immediately, and the user is notified with a warning. This process effectively reduces the risk of the user being deceived by a fraudulent call.

[0670] The following describes the processing flow.

[0671] Step 1:

[0672] The server monitors all incoming calls to the user's device and detects incoming calls from unregistered callers. This detection triggers the start of AI processing.

[0673] Step 2:

[0674] As soon as a call begins, the terminal collects voice data in real time and streams it to the server. During this process, the voice data is compressed to optimize bandwidth usage.

[0675] Step 3:

[0676] The server converts the received audio data into text using speech recognition technology. It also performs noise reduction processing to improve audio clarity.

[0677] Step 4:

[0678] The server compares the converted text against a database of scam talk patterns. Using natural language processing algorithms, it compares extracted key phrases and performs pattern recognition.

[0679] Step 5:

[0680] The server scores the likelihood of fraud. If the calculated score exceeds a threshold, it determines that there is a high probability of fraud and proceeds to the next step.

[0681] Step 6:

[0682] The server starts recording as soon as it determines that a call is highly likely to be a scam. This recording is saved as evidence and remains accessible to the user even after the call ends.

[0683] Step 7:

[0684] The server uses speech synthesis technology to generate a warning message and notifies both parties in the call in real time. This draws the listener's attention and puts pressure on the caller.

[0685] Step 8:

[0686] After the call ends, the server notifies the user of the fraud detection result and guides them to view the recorded data on a dedicated portal or application.

[0687] Step 9:

[0688] The server updates the fraudulent talk pattern database based on call data and judgment results. The AI ​​model continuously learns using newly collected data to improve its accuracy.

[0689] (Example 1)

[0690] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0691] In modern times, telephone fraud is becoming increasingly sophisticated year by year. This raises the risk of ordinary consumers unintentionally becoming victims of fraud, posing a significant security challenge, especially for the elderly and individuals with low IT literacy. Traditional methods such as manual registration restrictions and simple suspicious number blocking are insufficient; therefore, a comprehensive fraud prevention system employing new methods is needed.

[0692] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0693] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and streaming the audio data in real time using a data processing device, and means for converting the captured audio into text using speech recognition technology and comparing it with fraudulent talk patterns. This enables automatic detection of fraudulent activity and immediate response to it.

[0694] "Unregistered caller ID" refers to phone numbers or contact information that the user has not previously registered in the system.

[0695] "Means for detecting incoming calls" refers to a device or method that monitors telephone communications and identifies calls from unregistered callers.

[0696] "Means for capturing audio" refers to a device or mechanism for capturing audio signals generated during a telephone call and inputting them into a system.

[0697] A "data processing device" refers to a hardware or software component used for processing, converting, and analyzing audio signals.

[0698] "Means for streaming audio data" refers to a technology or method for transmitting acquired audio data in real time to other devices or systems.

[0699] "Speech recognition technology" refers to the technology that receives speech signals and converts them into text format that can be understood as human language.

[0700] "Fraudulent talk patterns" refer to typical conversational content and phrase patterns from past fraud cases, and are used as criteria for diagnosing the possibility of fraud.

[0701] A "pattern recognition algorithm" refers to a program or method that identifies certain regularities or features in data and performs classification or judgment based on specific conditions.

[0702] A "generative AI model" refers to an artificial intelligence model that learns new data patterns and uses them to generate data and make inferences.

[0703] By using the following configuration and technology in the implementation of this fraud prevention system, it is possible to smoothly and efficiently detect potential fraud and protect users.

[0704] First, the device receives an incoming call through the user's phone function. Upon receiving the call, the device's microphone is activated, and audio data from the call is collected in real time. This audio data is immediately streamed to the server via the communication network. Real-time data processing technologies such as WebRTC are used for streaming.

[0705] The server uses the received audio data to convert it to text using speech recognition software (e.g., Google Cloud Speech-to-Text). The transcribed data is then compared against a database containing recorded fraudulent talk patterns. Here, a pattern recognition algorithm using a natural language processing library (e.g., NLTK or Spacy) is applied to quantify the likelihood of fraud.

[0706] If a call is deemed highly likely to be fraudulent, the server immediately initiates a recording process. This stores the call content as evidence and generates a warning message for the user. This message generation uses synthesized speech technology (e.g., Amazon Polly). The warning is sent directly to both the caller and the recipient in real time during the call.

[0707] After the call, the server notifies the user of the details of the fraud assessment and the recorded data, prompting them to review it. The notification is sent via email or application push notification, and the user can review the content on a secure portal site or dedicated application. This information allows the user to review the call content and take steps to prevent becoming a victim of fraud.

[0708] Furthermore, the server automatically and regularly updates its database of fraudulent talk patterns and continuously learns to counter new fraudulent tactics using its generating AI model. This ensures that the system always reflects the latest fraud information and maintains a form that can adapt to ever-evolving fraudulent methods.

[0709] For example, if a call comes in from someone impersonating a "son" attempting to commit a wire transfer scam, the server detects phrases such as "accident" and "urgent transfer" and determines that it is highly likely to be a scam. At this stage, recording begins and a warning is sent to the user.

[0710] Example of a prompt:

[0711] "Is this call potentially a scam? The keywords included are 'accident' and 'urgent transfer'."

[0712] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0713] Step 1:

[0714] When a call is received, the terminal activates its built-in microphone and collects the call audio in real time. This audio data is the input. In other words, the terminal converts the audio signal into a digital format and sends it to the server in the appropriate format. The audio data, as output, is then moved to the next processing stage using real-time transmission technology.

[0715] Step 2:

[0716] The server receives audio data sent from the terminal. Next, a speech recognition component converts this audio data into text. The input data is an audio signal, and the output data is text in string format. Speech recognition technology is used to extract the content of the audio as textual information.

[0717] Step 3:

[0718] The server uses the converted text data to compare it against a database containing registered fraudulent talk patterns. Here, a pattern recognition algorithm is used to detect characteristic phrases of fraud and determine the likelihood of it being a scam. The input data is text, and the output data is a numerical rating of the likelihood of fraud.

[0719] Step 4:

[0720] If the likelihood of fraud exceeds a certain threshold, the server initiates the recording process. The input is the result of the fraud detection. As output, the recorded audio data is saved to the storage system. Simultaneously, a warning message is generated and sent to the user and caller using synthesized speech technology. This serves to alert the recipient.

[0721] Step 5:

[0722] After the call ends, the server notifies the user of the judgment result and the recorded data. The input is the recorded data and the judgment result, and the output is the notification information to the user. The notification is sent via email or application, and the user can check the content on a dedicated portal site or application.

[0723] Step 6:

[0724] The server automatically and regularly updates its database of fraudulent talk patterns and uses a generating AI model to flexibly adapt to new fraudulent methods. The input to this process is updated fraud information and historical data trends, while the output is an updated database and an improved model. In this way, the system can always operate with the most up-to-date fraud information.

[0725] (Application Example 1)

[0726] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0727] With the advancement of communication technology, fraudulent activities have diversified, and scams exploiting voice calls are on the rise. Such fraudulent activities are a serious problem, making it difficult for users to use voice calls with peace of mind. In particular, vulnerable groups such as the elderly are at risk of suffering significant losses due to the lack of adequate countermeasures against fraud. In this situation, there is a need for an efficient system that prevents fraudulent activities and provides users with a sense of security.

[0728] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0729] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for acquiring the audio of the incoming communication and processing the audio data in real time, and means for converting the acquired audio into text information and comparing it with fraudulent talk patterns. This enables the automatic detection and warning of potential fraudulent activity, allowing users to use voice calls with peace of mind.

[0730] "Incoming caller" refers to the identification information of the party that initiated the call or message in a communication method.

[0731] "Audio data" refers to information that represents sound in digital or analog format.

[0732] "Textual information" refers to text data obtained by converting data such as audio into a string of characters.

[0733] "Fraudulent behavior talk patterns" are a collection of characteristic voices and phrases extracted from past cases of fraudulent activity.

[0734] "Recording" is the process of saving certain data or information to prepare for future use.

[0735] A "warning" is a message issued to draw attention to risks or potential dangers.

[0736] A "warning audio message" is an audio notification used to warn or caution the sender or recipient when fraudulent activity is suspected.

[0737] A "user" is an entity that operates or utilizes a system, product, or service.

[0738] A "dedicated information terminal" is an electronic device designed specifically for a particular purpose or function.

[0739] An "application program" is software developed to achieve a specific function or purpose.

[0740] The system that realizes this invention mainly consists of three entities: a server, a terminal, and a user. The server is responsible for monitoring calls from unregistered incoming sources and processing those calls in real time. The terminal acquires the audio data of the call and sends it as a stream to the server. The server receives this and converts the audio into text data using the Google Cloud Speech-to-Text API. Based on this text data, it is compared against fraudulent talk patterns, and an AI model using TensorFlow or PyTorch analyzes and determines the possibility of fraud.

[0741] If the server detects a potential for fraud exceeding a threshold, the call is recorded, and a warning message is displayed on the user's device. Additionally, a cautionary voice message is generated and sent to the caller. This process allows the user to continue the call with confidence and later review the call content.

[0742] As a concrete example, consider a scenario where a user receives a phone call saying, "You've won a prize, please provide your bank account information." This system detects and analyzes phrases like "prize" and "bank account information," and if it determines that there is a high probability of fraudulent activity, it immediately starts recording and warns the user that "this may be a scam." An example of a prompt message would be, "Analyze the content of the call received by the user, analyze whether it contains the key phrases 'prize' and 'bank account information,' and issue a warning if it is suspicious."

[0743] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0744] Step 1:

[0745] The terminal detects an incoming call from an unregistered source and acquires audio data as soon as the call begins. This audio data serves as the input and is streamed to the server in real time.

[0746] Step 2:

[0747] The server converts the received audio data into text information using the Google Cloud Speech-to-Text API. This process outputs the audio data as text data.

[0748] Step 3:

[0749] The server compares the generated text data with a database of fraudulent talk patterns. A generative AI model using TensorFlow or PyTorch performs a probabilistic determination by comparing it with a database based on past cases. This process outputs a numerical result indicating the likelihood of fraudulent activity.

[0750] Step 4:

[0751] The server determines a threshold based on the quantified result, and if the likelihood of fraudulent activity exceeds the threshold, it starts recording the call. The input here is the judgment result, and the output is the start of the recording process.

[0752] Step 5:

[0753] A warning message based on the judgment result is displayed on the user's terminal. Warning information from the server is input and output to the user in the form of a notification.

[0754] Step 6:

[0755] The server generates a warning voice message and sends it to the caller. This process outputs a generated voice file, which is then played back by the caller to deliver the warning.

[0756] Step 7:

[0757] After the call ends, the server sends the judgment result and recorded data to the user's dedicated information terminal or application program for user review. In this step, the recorded data is the input, and it is output by being displayed on the user's review interface.

[0758] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0759] This invention combines an emotion engine with a fraud prevention system, providing a function to detect the user's emotions in real time during a call and more accurately determine the likelihood of fraud. This system monitors incoming calls from unregistered callers and analyzes the call content and the user's emotions to improve its ability to respond to fraud.

[0760] When a call comes in to a user's device, if it's from an unregistered caller, the server activates the emotion engine simultaneously with audio collection. Using speech recognition technology, it converts the audio into text and analyzes the user's emotional state from their voice patterns while comparing it against fraudulent talk patterns.

[0761] The device captures audio during a call with high accuracy and sends the necessary data to the server for the emotion engine. Factors such as tone, speed, and emphasis are considered.

[0762] The server uses an emotion engine to detect the user's emotions. The detected emotion data is fed back into the fraud possibility assessment and contributes to an overall decision. This result influences the fraud threshold determination and, if necessary, initiates recording or issues a warning.

[0763] If a user is emotionally distressed, the server will detect this and adjust the content and timing of the warning notification to help the user understand. For example, an additional warning in a calmer tone may be issued.

[0764] After the call ends, the server notifies the user of the detected emotion data and analysis results. This allows the user to understand their own emotional responses and use that information to improve future interactions.

[0765] Emotional data is stored as part of an ongoing learning process, contributing to the updating of the fraudulent talk pattern database and improving the accuracy of the AI ​​model. Using this learning process, the system continues to evolve and adapt to new fraudulent tactics and user emotional patterns.

[0766] For example, if someone claiming to be from a "financial institution" requests "account information verification," and the server detects that the user's voice indicates distress, the server will intensify the warning and provide an interaction prompting the user to acknowledge their emotions and exercise caution.

[0767] The following describes the processing flow.

[0768] Step 1:

[0769] The server monitors incoming calls to the user's terminal and detects calls from unregistered callers. This detection triggers the activation of voice data processing and the emotion engine.

[0770] Step 2:

[0771] As soon as a call starts, the device captures voice data with high precision and sends it to the server in stream format, including the voice parameters necessary for the emotion engine.

[0772] Step 3:

[0773] The server uses speech recognition technology to convert the audio data into text and compares it against a database of fraudulent talk patterns. In parallel, an emotion engine analyzes the tone, speed, emphasis, and other factors of the voice.

[0774] Step 4:

[0775] The server receives analysis results from the emotion engine and provides feedback to the fraud detection score if the user's emotions indicate anxiety or distress.

[0776] Step 5:

[0777] If the fraud probability score exceeds a threshold, the server will start recording and notify the user with a warning message tailored to their emotional state.

[0778] Step 6:

[0779] If a user shows emotional changes during a call, the server detects this in real time and dynamically adjusts the content and intensity of warning notifications to support user understanding.

[0780] Step 7:

[0781] After the call ends, the server notifies the user of the results, audio data, and sentiment analysis results. The user can then view this information through a dedicated portal or application.

[0782] Step 8:

[0783] The server updates the database using sentiment data and call history, and the AI ​​model continues to learn from the new data. This process allows the system to evolve and improve the accuracy of future fraud detections.

[0784] (Example 2)

[0785] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0786] Traditional fraud prevention systems relied on standard voice recognition and simple pattern matching for communications from unregistered sources, making it difficult to accurately determine the likelihood of fraud. Furthermore, they could not take into account the user's emotional state, resulting in inappropriate timing and content of warnings regarding fraudulent methods. Additionally, they struggled to respond quickly to the emergence of new fraud techniques.

[0787] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0788] In this invention, the server includes means for capturing transmitted audio and processing the audio information in real time, means for converting the captured audio into text and comparing it with fraud patterns, and means for analyzing the characteristics of the audio and determining the emotional state. This makes it possible to determine the possibility of fraud with high accuracy while taking into account the user's emotional state and to provide appropriate warnings and advice in real time. Furthermore, by regularly updating the information source for fraudulent talk patterns and improving accuracy through a learning device, it is possible to respond quickly to new fraudulent methods.

[0789] "Means for detecting communications" refers to a function that checks for and detects the presence of calls and messages from unregistered callers.

[0790] "Means for processing voice information" refers to a function that receives voice during a call and converts and manages it as digital data in real time.

[0791] "Means of converting to text" refers to a function that performs the process of converting audio data into text information using speech recognition technology.

[0792] "Methods for matching with fraud patterns" refers to a function that compares transcribed audio information with known fraudulent talk patterns and analyzes the degree of similarity.

[0793] "A means of analyzing voice characteristics and determining emotional state" refers to a function that analyzes features such as tone, speed, and emphasis of voice to identify the user's emotional response.

[0794] "Means of providing warnings and advice" refers to a function that alerts users and provides specific countermeasures when a potential scam is detected.

[0795] "Methods for regularly updating information sources" refers to the process of keeping fraud databases up-to-date and incorporating new fraud patterns.

[0796] A "learning device" is a technology necessary to train an AI model based on new information and continuously improve the system's discrimination accuracy.

[0797] This invention is a system aimed at preventing fraud, providing a function to analyze the user's emotions in real time during communication and determine the possibility of fraud. To achieve this, a system configuration in which a server and terminals are linked is necessary.

[0798] The server receives communications from the user's terminal. Specific software used includes speech recognition technology and generative AI models. The server transcribes the received audio data into text in real time and compares it against a fraud pattern database. During this process, features extracted from the audio data are used to activate an emotion engine and analyze the user's emotional state.

[0799] The terminal is responsible for capturing audio during calls with high precision. Hardware components include a microphone and an audio processing unit. The terminal converts the audio signal into digital data and sends it to the server. This allows the server to perform more accurate analysis.

[0800] As a concrete example, consider a scenario where someone claiming to be from a "financial institution" requests "confirmation of account information" from an unknown caller. In this situation, where the user is inevitably suspicious of fraud, the system detects the user's emotional distress and immediately intensifies the warning. The user is given calm voice advice such as, "Stay calm and do not carelessly provide personal information." This process is achieved by instructing the system to "detect keywords suggesting potential fraud through speech recognition and also perform emotional analysis of the user's emotional state."

[0801] In this system, emotional data is accumulated through a continuous learning process, and the fraudulent talk pattern database is also updated regularly. This allows the system to continuously adapt to new fraudulent tactics and changing user emotional patterns. By receiving this feedback, users can better understand their future unconscious reactions and take countermeasures.

[0802] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0803] Step 1:

[0804] The terminal detects incoming calls. Inputs include the caller's number and the call start signal. If the call is from an unregistered caller, the terminal immediately begins capturing audio data. The audio is captured by the microphone, noise is removed to produce clear audio data, and it is converted to a digital format for output.

[0805] Step 2:

[0806] The terminal transmits the captured audio data to the server in real time. During this process, acoustic characteristics such as tone, speed, and emphasis are input and sent directly to the server as digital data.

[0807] Step 3:

[0808] The server performs speech recognition based on the received audio data. Using a generative AI model, it converts speech to text, outputting text data from the audio data as input. During this process, advanced algorithms are employed to perform precise analysis of the audio data.

[0809] Step 4:

[0810] The server compares the transcribed text against a database of fraud patterns. It compares the text data received as input against known fraud patterns and outputs a degree of similarity based on the results. Based on the matching results, it makes an initial determination of the likelihood of fraud.

[0811] Step 5:

[0812] The server activates an emotion engine to analyze the user's emotions from the voice data. It takes voice characteristic data as input and outputs an emotional state (e.g., calm, agitated). The emotion analysis results are fed back into the fraud detection process.

[0813] Step 6:

[0814] The server comprehensively assesses the likelihood of fraud, issues a warning if necessary, and begins recording. In this step, it uses whether the assessed fraud likelihood score exceeds a threshold as a criterion, and adjusts the content and timing of the warning notification before sending it to the user. The output includes the warning content and stored recording data.

[0815] Step 7:

[0816] The server provides emotional feedback to the user after the call ends. Using the analyzed emotional data and fraud detection results as input, it outputs analysis results as appropriate and notifies the user via their terminal or a dedicated digital platform. Based on this, the user can consider future countermeasures.

[0817] Step 8:

[0818] The server feeds back the collected data to the learning device, updating the generated AI model. This process involves taking new fraud patterns and emotional patterns as input and outputting improvements in AI accuracy and system evolution.

[0819] (Application Example 2)

[0820] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0821] In recent years, telephone fraud tactics have become increasingly sophisticated, with a particular emphasis on appealing to emotions. Traditional fraud prevention systems rely solely on matching voice patterns and keywords, failing to consider the user's emotional state, which can lead to false positives and user confusion. Furthermore, users rarely have the opportunity to recognize their own emotional responses during a call, making it difficult for them to take preventative measures.

[0822] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0823] In this invention, the server includes means for detecting incoming calls from unregistered callers, means for capturing the audio of the incoming call and processing the audio data in real time, and means for analyzing the user's emotions from their voice patterns and determining the likelihood of fraud based on the emotional data. This makes it possible to more accurately assess the risk of fraud while recognizing the user's emotional response and to take warnings and risk avoidance actions.

[0824] "Unregistered caller" refers to an incoming call from a phone number that is not pre-registered on the user's device or system.

[0825] "Audio data processing" refers to a series of operations that analyze or transform audio acquired during a call for a specific purpose.

[0826] "Converting speech to text" is the process of automatically replacing spoken information generated during a phone call with written text.

[0827] A "fraudulent talk pattern" refers to the characteristics of audio data that include conversational content or specific phrases intended for fraudulent purposes.

[0828] "Emotional data" refers to information that quantifies or represents the emotional tendencies and reactions that a user exhibits during a call.

[0829] "Notifying a warning" refers to providing users with information to inform them of the risks and encourage them to take safety measures when fraud is suspected.

[0830] A "learning model" is an algorithm or system that uses machine learning to recognize and predict specific data patterns.

[0831] "Emotional feedback" is an information provision process that informs users about the content and impact of their emotional responses during a call.

[0832] This invention is a system that uses the user's communication device to analyze the possibility of fraud in real time on the server side and issue warnings based on sentiment data. To implement this system, the user's terminal first detects an incoming call from an unregistered source. When a voice call begins, the terminal captures the audio during the call with high accuracy and transmits it to the server in real time.

[0833] The server uses speech recognition technologies such as the Google Cloud Speech-to-Text API to convert received audio into text data. This text data is compared against a database of fraudulent talk patterns and analyzed to determine the likelihood of fraud. Simultaneously, the server uses the Python transformers library to analyze the user's emotional state from the transcribed conversation. The emotional data obtained in this process is fed back into the judgment process.

[0834] If the server determines that a user is facing a risk of fraud, it sends a warning notification to the user's device via the Flutter framework, and the audio is recorded. This allows the user to understand the risks during the call and take appropriate action. After the call ends, the server also provides the user with emotional feedback and encourages self-analysis of the past call through a dedicated application.

[0835] For example, if a user is approached with a fraudulent financial transaction offer, and the audio matches a scammer's conversation pattern, and the system detects signs of distress in the user, it will immediately issue a warning and begin recording. This allows the user to gain knowledge about the fraudulent activity and strengthen their preparations for the next steps.

[0836] An example of a prompt to a generative AI model is: "Assess the likelihood of fraud based on the following sentiment data and text in response to the caller's call, and output the result. Sentiment data: agitated, frightened. Text: 'Please provide your account information.'" By using this prompt, the AI ​​model can gain a deeper understanding of the caller's intent and effectively assess the risk of fraud.

[0837] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0838] Step 1:

[0839] The terminal detects incoming calls from unregistered callers. At this time, it receives the caller's number information as input, and if it does not exist in the internal registration database, it determines the caller is unknown and activates the system.

[0840] Step 2:

[0841] The terminal captures audio during a call and sends it to the server with high accuracy. It receives audio data as input, encodes it, and sends it to the server in real time. Here, noise cancellation technology is used to maintain clear audio data.

[0842] Step 3:

[0843] The server uses the Google Cloud Speech-to-Text API to convert audio data into text. The server receives audio data as input and applies speech recognition technology to generate text output. This output text forms the basis for the next analysis step.

[0844] Step 4:

[0845] The server uses text data to match it against scam talk patterns. Here, it takes text data as input and compares it with a scam database to perform data calculations that identify phrases that match predefined scam patterns.

[0846] Step 5:

[0847] The server uses the transformers library to analyze the user's emotional state from the text. This process quantifies the emotional data extracted from the input text and classifies it into categories such as positive, negative, and neutral. The calculated emotional data is then used in the next decision step.

[0848] Step 6:

[0849] The server determines the risk of fraud based on the degree of match with fraud patterns and the user's sentiment data. It receives sentiment data and fraud match scores as input, compares them to a threshold, and outputs whether a warning is necessary. Based on this result, it decides whether a warning is sent to the user.

[0850] Step 7:

[0851] If the server determines that the level has exceeded the danger threshold, it sends a warning notification to the terminal and starts recording. At this point, it uses the final risk assessment result as input to generate a warning message, displays the warning on the terminal, and starts the process of saving the audio data.

[0852] Step 8:

[0853] After the call ends, the server generates emotional feedback and notifies the user. The emotional data collected during the call is aggregated and converted into a feedback format that is easy for the user to understand. This result is then communicated to the user and used to improve future fraud prevention efforts.

[0854] 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 controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0855] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0856] 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 this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[0857] Furthermore, the emotion identification model 59, acting 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 a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0858] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[0859] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[0860] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[0861] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[0862] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is 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 the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[0863] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[0864] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[0865] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[0866] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0867] 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.

[0868] Furthermore, it is not necessary to store the entirety 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 the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[0869] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[0870] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of 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). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[0871] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[0872] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[0873] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[0874] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0875] The following is further disclosed regarding the embodiments described above.

[0876] (Claim 1)

[0877] A means of detecting incoming calls from unregistered callers,

[0878] A means of capturing the audio of an incoming call and processing the audio data in real time,

[0879] A method for converting captured audio into text and comparing it with fraudulent talk patterns,

[0880] A system that includes means to determine the possibility of fraud and to start recording and notify a warning if a threshold is exceeded.

[0881] (Claim 2)

[0882] The system according to claim 1, comprising: regularly updating a database of fraudulent talk patterns and maintaining a learning model that can adapt to the evolution of fraudulent methods.

[0883] (Claim 3)

[0884] The system according to claim 1, wherein the judgment result and recorded data are notified to the user and can be viewed through a dedicated portal site or application.

[0885] "Example 1"

[0886] (Claim 1)

[0887] A means of detecting incoming calls from unregistered callers,

[0888] A means for capturing the audio of an incoming call and streaming the audio data in real time using a data processing device,

[0889] A method for converting captured audio into text using speech recognition technology and comparing it with fraudulent talk patterns,

[0890] A means of applying a pattern recognition algorithm to determine the possibility of fraud, starting recording if a threshold is exceeded, and notifying a warning using a warning generation device,

[0891] A means of notifying the user of the judgment result and recording audio data,

[0892] A system that includes means to regularly update the database and utilize generative AI models to respond to new fraudulent methods.

[0893] (Claim 2)

[0894] The system according to claim 1, which updates a database of fraudulent talk patterns and uses a learning model to respond to the evolution of fraudulent methods.

[0895] (Claim 3)

[0896] The system according to claim 1, which notifies the user of the judgment result and recorded data, and allows them to confirm it via a communication network.

[0897] "Application Example 1"

[0898] (Claim 1)

[0899] A means of detecting incoming calls from unregistered callers,

[0900] A means for acquiring the audio of an incoming call and processing the audio data in real time,

[0901] A means of converting acquired audio into text information and comparing it with fraudulent talk patterns,

[0902] A means to determine the possibility of fraudulent activity, start recording and notify a warning if a threshold is exceeded,

[0903] A means of quickly displaying the results on the user's device,

[0904] A system that includes a means of generating and outputting a warning voice message to the caller if fraudulent activity is detected.

[0905] (Claim 2)

[0906] The system according to claim 1, which periodically updates a database of fraudulent talk patterns and maintains a learning model that responds to the evolution of fraudulent methods.

[0907] (Claim 3)

[0908] The system according to claim 1, wherein the judgment result and recorded data are notified to the user and can be confirmed through a dedicated information terminal or application program.

[0909] "Example 2 of combining an emotion engine"

[0910] (Claim 1)

[0911] A means of detecting communications from unregistered sources,

[0912] A means for capturing transmitted audio and processing the audio information in real time,

[0913] A method for converting captured audio into text and comparing it with fraud patterns,

[0914] A means of analyzing the characteristics of speech and determining emotional state,

[0915] A system that includes means to determine the possibility of fraud and to start recording and notify a warning if a threshold is exceeded.

[0916] (Claim 2)

[0917] The system according to claim 1, which regularly updates its source of information on fraudulent talk patterns, maintains a learning device that can adapt to the evolution of fraudulent methods, and improves accuracy by feeding back the results of sentiment analysis.

[0918] (Claim 3)

[0919] The system according to claim 1, which notifies the user of the communication judgment results and recorded information, makes them verifiable through a dedicated digital platform or application, and provides feedback based on sentiment analysis.

[0920] "Application example 2 when combining with an emotional engine"

[0921] (Claim 1)

[0922] A means of detecting incoming calls from unregistered callers,

[0923] A means of capturing the audio of an incoming call and processing the audio data in real time,

[0924] A method for converting captured audio into text and comparing it with fraudulent talk patterns,

[0925] A method for analyzing emotions from a user's voice patterns and determining the likelihood of fraud based on emotional data,

[0926] A system that includes means for starting recording and issuing a warning when a threshold is exceeded.

[0927] (Claim 2)

[0928] The system according to claim 1, comprising: regularly updating a database of fraudulent talk patterns; maintaining a learning model that can adapt to the evolution of fraudulent methods; and continuously learning user sentiment data to improve the accuracy of the AI ​​model.

[0929] (Claim 3)

[0930] The system according to claim 1, which notifies the user of the judgment result and recorded data, provides emotional feedback through a dedicated portal site or application, and allows the user to review it so that it can be used to improve future countermeasures. [Explanation of symbols]

[0931] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of detecting incoming calls from unregistered callers, A means for acquiring the audio of an incoming call and processing the audio data in real time, A means of converting acquired audio into text information and comparing it with fraudulent talk patterns, A means to determine the possibility of fraudulent activity, start recording and notify a warning if a threshold is exceeded, A means of quickly displaying the results on the user's device, A system that includes a means of generating and outputting a warning voice message to the caller if fraudulent activity is detected.

2. The system according to claim 1, which periodically updates a database of fraudulent talk patterns and maintains a learning model that can respond to the evolution of fraudulent methods.

3. The system according to claim 1, wherein the judgment result and recorded data are notified to the user and can be confirmed through a dedicated information terminal or application program.

Images