System

Abstract

An object of a system according to an embodiment is to detect a fraud targeting an elderly person in real time and to quickly respond to the fraud.SOLUTION: A system includes a voice recognition unit, an analysis unit, an alert unit, and an intervention unit. The voice recognition unit recognizes a voice conversation in real time. The analysis unit analyzes the voice recognized by the voice recognition unit. The alert unit issues an alert when a risk of fraud is detected based on the result analyzed by the analysis unit. The intervener intervenes in the conversation when the alert is issued by the alerter.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] Conventional technology has the problem of making it difficult to detect fraud targeting the elderly in real time and respond quickly.

[0005] The system according to the embodiment aims to detect fraud targeting the elderly in real time and respond quickly. [Means for solving the problem]

[0006] The system according to the embodiment includes a voice recognition unit, an analysis unit, an alert unit, and an intervention unit. The voice recognition unit recognizes voice conversations in real time. The analysis unit analyzes the voice recognized by the voice recognition unit. The alert unit issues an alert when a risk of fraud is detected based on the results of the analysis by the analysis unit. The intervention unit intervenes in the conversation when an alert is issued by the alert unit. [Effects of the Invention]

[0007] The system according to the embodiment can detect fraud targeting elderly people in real time and respond quickly. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. DETAILED DESCRIPTION OF THE INVENTION

[0009] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0010] First, the terms used in the following description will be explained.

[0011] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, the processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (Tensor Processing Unit).

[0012] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0013] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0014] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), and Bluetooth (registered trademark).

[0015] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0016] [First embodiment] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0017] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0018] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0019] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0020] The reception device 38 includes a touch panel 38A and a microphone 38B, and receives user input. The touch panel 38A detects contact with a pointer (for example, a pen or a finger) to receive user input by the touch of the pointer. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 (see FIG. 2) acquires the data indicating the user input.

[0021] Output device 40 includes a display 40A and a speaker 40B, and presents data to a user by outputting the data in a form of expression that the user can perceive (e.g., audio and / or text). Display 40A displays visible information such as text and images in accordance with instructions from processor 46. Speaker 40B outputs audio in accordance with instructions from processor 46. Camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0022] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0023] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0024] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0025] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate a user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotion, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, the estimation and prediction of emotion also includes, for example, emotion analysis.

[0026] In the smart device 14, the specific processing is performed by the processor 46. The storage 50 stores a specific processing program 60. The specific processing program 60 is used together with the specific processing program 56 by the data processing system 10. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as the control unit 46A in accordance with the specific processing program 60 executed on the RAM 48. Note that the smart device 14 has a data generation model and an emotion identification model similar to the data generation model 58 and the emotion identification model 59, and can also perform processing similar to that of the specific processing unit 290 using these models.

[0027] Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains a processing result (prediction result, etc.) using the data generation model 58 by communicating with the server device having the data generation model 58. Furthermore, the data processing device 12 may be a server device, or may be a terminal device owned by a user (e.g., a mobile phone, a robot, a home appliance, etc.). Next, an example of processing by the data processing system 10 according to the first embodiment will be described.

[0028] (Example 1) A fraud prevention system according to an embodiment of the present invention is a system for preventing "it's my son" frauds committed by elderly people. This fraud prevention system recognizes voice conversations in real time, analyzes their content using AI, issues an alert if there is a possibility of fraud, and intervenes in the conversation as necessary. For example, the fraud prevention system recognizes voice conversations in real time. For example, the fraud prevention system converts what an elderly person is saying on the phone into text in real time, and then analyzes the text using AI. The fraud prevention system then determines whether there is a possibility of fraud based on the text analyzed by the AI. For example, if the text contains phrases such as "please transfer money" or "please handle this quickly," the fraud prevention system determines that there is a high possibility of fraud. Next, if the fraud prevention system determines that there is a possibility of fraud, it issues an alert. For example, it sends a notification to the elderly person's smartphone or notifies their family by email. Furthermore, the fraud prevention system intervenes in the conversation as necessary. For example, the AI ​​automatically interrupts the conversation and warns them that "this conversation may be fraudulent." Family members can also remotely participate in the conversation and provide support to the elderly. In this way, the fraud prevention system can prevent "it's my son" frauds committed by elderly people. This allows the fraud prevention system to prevent "it's me" scams against the elderly. For example, if a fraudulent phone call is received, the fraud prevention system will automatically respond, preventing the elderly from becoming victims. The elderly can then live their daily lives with peace of mind, and their families can watch over them with peace of mind.

[0029] The fraud prevention system according to the embodiment includes a voice recognition unit, an analysis unit, an alert unit, and an intervention unit. The voice recognition unit recognizes voice conversations in real time. The voice recognition unit converts voice into text, for example, using a deep learning model. The voice recognition unit converts, for example, what an elderly person is saying on the phone into text in real time. The voice recognition unit converts voice into text with high accuracy, for example, using a deep learning model. The analysis unit analyzes the voice recognized by the voice recognition unit. The analysis unit determines the possibility of fraud, for example, using a model trained on past fraud cases. The analysis unit determines the possibility of fraud if the voice contains phrases such as "please transfer money" or "please handle this quickly." The analysis unit determines the possibility of fraud with high accuracy, for example, using a model trained on past fraud cases. The alert unit issues an alert if a risk of fraud is detected based on the results of the analysis by the analysis unit. The alert unit issues a warning, for example, via smartphone notification, email, or voice alert. The alert unit, for example, sends a notification to the elderly person's smartphone or notifies family members by email. The alert unit can quickly notify the user of the possibility of fraud by, for example, issuing a smartphone notification, email, or voice warning. The intervention unit intervenes in the conversation when the alert unit issues an alert. For example, the intervention unit automatically interrupts the conversation using AI to warn the user that "this conversation may be fraudulent." For example, the intervention unit allows a family member to remotely participate in the conversation and support the elderly. For example, the intervention unit automatically interrupts the conversation using AI to warn the user of the possibility of fraud, thereby reducing the risk of elderly people falling victim to fraud. As a result, the fraud prevention system according to the embodiment can prevent "it's my son" frauds committed by elderly people.

[0030] The speech recognition unit can convert speech into text using a deep learning model. Examples of deep learning models include, but are not limited to, a convolutional neural network (CNN), a recurrent neural network (RNN), and a transformer. The speech recognition unit can, for example, use a CNN to analyze speech data and convert it into text. The speech recognition unit can also use an RNN to analyze temporal dependencies in speech data and convert it into text. The speech recognition unit can also use a Transformer to analyze speech data with high accuracy and convert it into text. For example, the speech recognition unit can use a CNN to extract features of speech data and convert it into text. The RNN analyzes temporal dependencies in speech data and converts continuous speech data into text. The Transformer analyzes speech data with high accuracy and converts it into text taking context into account. In this way, the deep learning model can convert speech into text with high accuracy. Some or all of the above-mentioned processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can convert speech to text using an AI model that uses a deep learning model to input speech data and outputs text data.

[0031] The analysis unit can determine the likelihood of fraud using a model trained on past fraud cases. Examples of models trained on past fraud cases include, but are not limited to, fraud case datasets and supervised learning. The analysis unit, for example, trains a model using a fraud case dataset to determine the likelihood of fraud. The analysis unit can also use supervised learning to learn fraud patterns and determine the likelihood of fraud. The analysis unit can also combine a fraud case dataset and supervised learning to determine the likelihood of fraud with high accuracy. For example, the analysis unit trains a model using a fraud case dataset to determine the likelihood of fraud when specific keywords or phrases are included. Supervised learning learns fraud patterns and enables adaptation to new fraud methods. Combining a fraud case dataset and supervised learning can determine the likelihood of fraud with high accuracy. As a result, using a model trained on past fraud cases can determine the likelihood of fraud with high accuracy. Some or all of the above-described processing in the analysis unit may be performed, for example, using AI or without AI. For example, the analysis unit can determine the likelihood of fraud using an AI model that takes a fraud case dataset as input and outputs the likelihood of fraud.

[0032] The alert unit can issue a smartphone notification, email, or voice alert. Examples of smartphone notifications include, but are not limited to, push notifications, SMS, and in-app notifications. The alert unit can send an alert to the elderly person's smartphone using, for example, push notifications. The alert unit can also send an alert to the elderly person's smartphone using SMS. The alert unit can also send an alert to the elderly person's smartphone using an in-app notification. Examples of emails include, but are not limited to, HTML emails and text emails. The alert unit can send an alert to family members using, for example, HTML emails. The alert unit can also send an alert to family members using text emails. Examples of voice alerts include, but are not limited to, voice synthesis and recorded voices. The alert unit can send an alert to the elderly person using, for example, voice synthesis. The alert unit can also send an alert to the elderly person using recorded voices. This allows the user to be quickly notified of a possible fraud by issuing a smartphone notification, email, or voice alert. Some or all of the above-described processing by the alert unit can be performed using, for example, AI, or without AI. For example, the alerting unit can use AI models to issue alerts via smartphone notifications, emails, and voice warnings when potential fraud is detected.

[0033] The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud." Methods for the AI ​​to automatically interrupt a conversation include, but are not limited to, when a specific keyword appears or after a certain period of time has passed. For example, the intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a specific keyword appears. The intervention unit can also automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," after a certain period of time has passed. The intervention unit can also automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when the AI ​​analyzes the content of the conversation and determines that there is a high risk of fraud. For example, the intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a specific keyword appears. The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a certain period of time has passed. The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when the AI ​​analyzes the content of the conversation and determines that there is a high risk of fraud. In this way, the AI ​​can automatically interrupt the conversation and warn, reducing the risk of elderly people falling victim to fraud. Some or all of the above-described processing by the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may interrupt a conversation using an AI model that takes the content of the conversation as input and outputs a warning about the risk of fraud.

[0034] The intervention unit can support the elderly by having a family member remotely participate in the conversation. Methods for remotely participating in the conversation include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using a video call. The intervention unit can also support the elderly by having a family member remotely participate in the conversation using a voice call. The intervention unit can also support the elderly by having a family member remotely participate in the conversation using a chat. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using a video call and warn of the risk of fraud. The family can support the elderly by having a family member remotely participate in the conversation using a voice call. The family can support the elderly by having a family member remotely participate in the conversation using a chat. In this way, the elderly can be supported by having a family member remotely participate in the conversation and the risk of fraud can be further reduced. Some or all of the above-described processing in the intervention unit can be performed using, for example, AI, or without AI. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using an AI model that sets up a video call, voice call, or chat for remotely participating in the conversation.

[0035] The voice recognition unit may add a filtering function that automatically removes background noise during voice recognition. Methods for automatically removing background noise include, but are not limited to, noise canceling algorithms and filtering techniques. The voice recognition unit may, for example, use a noise canceling algorithm to automatically remove background noise during voice recognition. The voice recognition unit may also use filtering techniques to remove noise in a specific frequency band during voice recognition. The voice recognition unit may also learn environmental sounds and automatically remove noise of a specific pattern. For example, the voice recognition unit may use a noise canceling algorithm to analyze ambient noise in real time and remove noise. Filtering techniques may be used to remove noise in a specific frequency band and recognize only speech. Environmental sounds may be learned and noise of a specific pattern may be automatically removed. This automatically removing background noise improves the accuracy of voice recognition. Some or all of the above-described processing in the voice recognition unit may be performed using, for example, AI, or may be performed without AI. For example, the voice recognition unit may automatically remove background noise using an AI model that receives background noise as input and outputs noise-removed voice data.

[0036] During speech recognition, the speech recognition unit can learn the characteristics of a speaker's voice and use an individually optimized speech recognition model. Methods for learning the characteristics of a speaker's voice include, but are not limited to, extraction of acoustic features and training of a machine learning model. For example, the speech recognition unit can learn the tone and pitch of a speaker's voice using the extraction of acoustic features and generate an individually optimized speech recognition model. The speech recognition unit can also learn the pronunciation habits and speed of a speaker using training of a machine learning model and use an individually optimized speech recognition model. The speech recognition unit can also continuously learn the characteristics of a speaker's voice and update the speech recognition model. For example, the speech recognition unit can learn the tone and pitch of a speaker using the extraction of acoustic features and generate an individually optimized speech recognition model. The speech recognition unit can also learn the pronunciation habits and speed of a speaker using training of a machine learning model and use an individually optimized speech recognition model. The speech recognition unit continuously learns the characteristics of a speaker's voice and updates the speech recognition model. As a result, by learning the characteristics of a speaker's voice and using an individually optimized speech recognition model, the accuracy of speech recognition is improved. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that receives input of feature data of a speaker's voice and outputs an individually optimized speech recognition model.

[0037] The speech recognition unit may apply an algorithm that takes into account the context of a conversation during speech recognition to reduce misrecognition. Methods for taking into account the context of a conversation include, but are not limited to, analysis of preceding and following utterances and context-based algorithms. For example, the speech recognition unit may apply an algorithm that takes into account the context of a conversation and reduces misrecognition. The speech recognition unit may also use a context-based algorithm to analyze the content of a conversation in real time and improve the accuracy of speech recognition based on the context. The speech recognition unit may also learn the topic of a conversation and reduce misrecognition based on a specific context. For example, the speech recognition unit may use an analysis of preceding and following utterances to apply an algorithm that takes into account the context of a conversation and reduces misrecognition. The context-based algorithm may analyze the content of a conversation in real time and improve the accuracy of speech recognition based on the context. The topic of a conversation may be learned and misrecognition may be reduced based on a specific context. Thus, misrecognition can be reduced by taking the context of a conversation into account. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI or without AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that takes conversational context data as input and outputs an algorithm that reduces misrecognition.

[0038] The speech recognition unit can improve the recognition accuracy by taking into account the speaker's geographical accent or dialect during speech recognition. Methods for taking into account geographical accents or dialects include, but are not limited to, learning regional speech data and using accent conversion algorithms. For example, the speech recognition unit learns the speaker's geographical accent using regional speech data learning and reflects it in a speech recognition model. The speech recognition unit can also analyze the speaker's dialect using an accent conversion algorithm to customize the speech recognition model. The speech recognition unit can also apply an algorithm that improves the speech recognition accuracy based on the geographical accent or dialect. For example, the speech recognition unit learns regional speech data, learns the speaker's geographical accent, and reflects it in a speech recognition model. The speech recognition unit analyzes the speaker's dialect using an accent conversion algorithm and customizes the speech recognition model. An algorithm that improves the speech recognition accuracy based on the geographical accent or dialect is applied. As a result, the accuracy of speech recognition is improved by taking into account the speaker's geographical accent or dialect. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or without AI. For example, the speech recognition unit can use AI models that take geographic accent and dialect data as input and customize the speech recognition model to improve speech recognition accuracy.

[0039] The speech recognition unit can customize a speech recognition model based on the age and gender of the speaker during speech recognition. Methods of customization based on age and gender include, but are not limited to, speech models for each age group and speech features for each gender. For example, the speech recognition unit customizes the speech recognition model based on the speaker's age using a speech model for each age group. The speech recognition unit can also optimize the speech recognition model based on the speaker's gender using speech features for each gender. The speech recognition unit can also apply an algorithm that improves speech recognition accuracy by taking age and gender into consideration. For example, the speech recognition unit customizes the speech recognition model based on the speaker's age using a speech model for each age group. The speech recognition unit optimizes the speech recognition model based on the speaker's gender using speech features for each gender. The speech recognition unit applies an algorithm that improves speech recognition accuracy by taking age and gender into consideration. Customizing the speech recognition model based on the speaker's age and gender improves speech recognition accuracy. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that takes data such as the speaker's age and gender as input and customizes the speech recognition model.

[0040] During speech recognition, the speech recognition unit can improve recognition accuracy by referring to a speaker's past conversation history. Methods for referring to the past conversation history include, but are not limited to, storing and searching a conversation log and analyzing historical data. For example, the speech recognition unit learns the speaker's past conversation history using the storing and searching of a conversation log and reflects it in a speech recognition model. The speech recognition unit can also analyze the content of the speaker's past conversations using analysis of historical data to improve speech recognition accuracy. The speech recognition unit can also combine the storing and searching of a conversation log and the analysis of historical data to learn the speaker's past conversation patterns and apply an algorithm that reduces recognition errors. For example, the speech recognition unit learns the speaker's past conversation history using the storing and searching of a conversation log and reflects it in a speech recognition model. The content of the speaker's past conversations using analysis of historical data can be analyzed to improve speech recognition accuracy. By combining the storing and searching of a conversation log and the analysis of historical data to learn the speaker's past conversation patterns and apply an algorithm that reduces recognition errors, speech recognition accuracy is improved. In this way, the speech recognition accuracy is improved by referring to the speaker's past conversation history. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit may use an AI model that uses the speaker's past conversation history data as input and customizes the speech recognition model to improve the accuracy of speech recognition.

[0041] During analysis, the analysis unit can apply an algorithm that takes into account the context of the conversation to more accurately determine the likelihood of fraud. Methods that take into account the context of the conversation include, but are not limited to, analysis of surrounding utterances and context-based algorithms. For example, the analysis unit applies an algorithm that takes into account the context of the conversation and more accurately determines the likelihood of fraud. The analysis unit can also use a context-based algorithm to analyze the content of the conversation in real time and determine the likelihood of fraud based on the context. The analysis unit can also learn the topic of the conversation and determine the likelihood of fraud based on the specific context. For example, the analysis unit applies an algorithm that takes into account the context of the conversation and more accurately determines the likelihood of fraud by using analysis of surrounding utterances. The context-based algorithm analyzes the content of the conversation in real time and determines the likelihood of fraud based on the context. The topic of the conversation is learned and the likelihood of fraud is determined based on the specific context. In this way, by taking the context of the conversation into account, the likelihood of fraud can be more accurately determined. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI or without AI. For example, the analysis unit can more accurately determine the likelihood of fraud by using an AI model that takes conversation context data as input and outputs an algorithm that determines the likelihood of fraud.

[0042] During analysis, the analysis unit can learn not only past fraud cases but also the latest fraud techniques to improve analysis accuracy. Methods for learning the latest fraud techniques include, but are not limited to, collecting and learning from the latest fraud datasets and online learning algorithms. For example, the analysis unit can use the collection and learning of the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques, thereby improving analysis accuracy. The analysis unit can also use an online learning algorithm to learn the latest fraud techniques in real time to improve analysis accuracy. The analysis unit can also combine past fraud cases with the latest fraud techniques to update the analysis model. For example, the analysis unit can use the collection and learning of the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques, thereby improving analysis accuracy. The online learning algorithm can learn the latest fraud techniques in real time to improve analysis accuracy. The analysis model is updated by combining past fraud cases with the latest fraud techniques. In this way, by learning the latest fraud techniques, analysis accuracy is improved. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can improve analysis accuracy by using an AI model that takes the latest fraud data set as input and updates the analysis model.

[0043] During analysis, the analysis unit can analyze the tone and speed of speech to determine the possibility of fraud. Methods for analyzing the tone and speed of speech include, but are not limited to, extraction of speech features and tone analysis algorithms. For example, the analysis unit can analyze the tone of speech using extraction of speech features to determine the possibility of fraud. The analysis unit can also analyze the tone of speech in real time using a tone analysis algorithm to determine the possibility of fraud. The analysis unit can also analyze the speed of speech in real time to determine the possibility of fraud. For example, the analysis unit can analyze the tone of speech using extraction of speech features to determine the possibility of fraud. The analysis unit can analyze the tone of speech in real time using a tone analysis algorithm to determine the possibility of fraud. The analysis unit can analyze the speed of speech in real time to determine the possibility of fraud. In this way, by analyzing the tone and speed of speech, the possibility of fraud can be more accurately determined. Some or all of the above-described processing in the analysis unit may be performed using, for example, AI, or may be performed without AI. For example, the analysis unit can input data on the tone and speed of speech and use an AI model that determines the possibility of fraud to determine the possibility of fraud.

[0044] During analysis, the analysis unit can determine the possibility of fraud by analyzing not only the content of the conversation but also the speaker's tone of voice and emotion. Methods for analyzing tone of voice and emotion include, but are not limited to, extraction of voice features and emotion analysis algorithms. For example, the analysis unit can analyze the content of the conversation using extraction of voice features and determine the possibility of fraud by taking into account the speaker's tone of voice. The analysis unit can also use an emotion analysis algorithm to analyze the speaker's emotion in real time and determine the possibility of fraud. The analysis unit can also combine the content of the conversation and the speaker's emotion to more accurately determine the possibility of fraud. For example, the analysis unit can analyze the content of the conversation using extraction of voice features and determine the possibility of fraud by taking into account the speaker's tone of voice. The emotion analysis algorithm can analyze the speaker's emotion in real time and determine the possibility of fraud. The combination of the content of the conversation and the speaker's emotion can more accurately determine the possibility of fraud. In this way, by analyzing the speaker's tone of voice and emotion, the possibility of fraud can be more accurately determined. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can input data on the content of the conversation, tone of voice, and emotions, and use an AI model that determines the possibility of fraud to determine the possibility of fraud.

[0045] During analysis, the analysis unit can determine the possibility of fraud by referring to the speaker's past behavioral history. Methods of referring to past behavioral history include, but are not limited to, storing and searching behavior logs and analyzing history data. For example, the analysis unit can learn the speaker's past behavioral history using the storing and searching of behavior logs and determine the possibility of fraud. The analysis unit can also analyze the speaker's past behavioral patterns using the analysis of history data to predict the possibility of fraud. The analysis unit can also combine the storing and searching of behavior logs and the analysis of history data to refer to the speaker's past behavioral history in real time and determine the possibility of fraud. For example, the analysis unit can use the storing and searching of behavior logs and the analysis of history data to learn the speaker's past behavioral history and determine the possibility of fraud. The analysis unit can use the storing and searching of behavior logs and the analysis of history data to analyze the speaker's past behavioral patterns and predict the possibility of fraud. The analysis unit can combine the storing and searching of behavior logs and the analysis of history data to refer to the speaker's past behavioral history in real time and determine the possibility of fraud. By doing so, the possibility of fraud can be more accurately determined by referring to the speaker's past behavioral history. Some or all of the above-described processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can input the speaker's past behavioral history data and determine the possibility of fraud using an AI model that determines the possibility of fraud.

[0046] During analysis, the analysis unit can determine the possibility of fraud by taking into account the speaker's geographic location information. Methods for taking into account geographic location information include, but are not limited to, the use of GPS data and a location-based analysis algorithm. For example, the analysis unit can use GPS data to analyze the speaker's geographic location information and determine the possibility of fraud. The analysis unit can also use a location-based analysis algorithm to refer to the speaker's location information in real time and predict the possibility of fraud. The analysis unit can also combine the use of GPS data and a location-based analysis algorithm to determine the possibility of fraud by taking into account the speaker's geographic location information. For example, the analysis unit can use GPS data to analyze the speaker's geographic location information and determine the possibility of fraud. The location-based analysis algorithm can refer to the speaker's location information in real time and predict the possibility of fraud. The use of GPS data and a location-based analysis algorithm can be combined to determine the possibility of fraud by taking into account the speaker's geographic location information. This allows for a more accurate determination of the possibility of fraud by taking into account the speaker's geographic location information. Some or all of the above-described processing in the analysis unit can be performed, for example, using AI or without AI. For example, the analysis unit can use an AI model that takes the speaker's geographic location data as input and determines the likelihood of fraud.

[0047] When issuing an alert, the alert unit can combine multiple notification means to issue the alert. Examples of multiple notification means include, but are not limited to, smartphone notifications, emails, and audio alerts. For example, the alert unit can simultaneously send a smartphone notification and an email to quickly notify the user of the alert. The alert unit can also combine an audio alert and a smartphone notification to quickly notify the user of the alert both visually and audibly. The alert unit can also combine an email and an audio alert to reliably notify the user of the alert. For example, the alert unit can simultaneously send a smartphone notification and an email to quickly notify the user of the alert. The audio alert and a smartphone notification can be combined to quickly notify the user of the alert both visually and audibly. The email and audio alert can be combined to reliably notify the user of the alert. In this way, by combining multiple notification means, the alert can be quickly and reliably notified of the alert. Some or all of the above-described processing in the alert unit may be performed, for example, using AI, or may be performed without AI. For example, the alert unit can issue an alert using an AI model that receives the alert content as input and outputs an alert using multiple notification means.

[0048] When issuing an alert, the alert unit can select the most appropriate alert method by referring to past alert history. Methods for referencing past alert history include, but are not limited to, saving and searching alert logs and analyzing historical data. The alert unit, for example, analyzes past alert history by saving and searching alert logs and selects the most effective notification method. The alert unit can also prioritize alert methods to which the user has responded in the past by analyzing historical data. The alert unit can also learn user response patterns by combining saving and searching alert logs and analyzing historical data and propose the most appropriate alert method. For example, the alert unit analyzes past alert history by saving and searching alert logs and selects the most effective notification method. Prioritizes alert methods to which the user has responded in the past by analyzing historical data. Learns user response patterns by combining saving and searching alert logs and analyzing historical data and proposes the most appropriate alert method. In this way, the most appropriate alert method can be selected by referring to past alert history. Some or all of the above-described processing in the alert unit may be performed, for example, using AI or without AI. For example, the alert unit can select an alert method using an AI model that inputs past alert history data and selects the optimal alert method.

[0049] The alert unit can customize the content of the alert when issuing an alert and notify the user in an optimal manner. Methods for customizing the content of the alert include, but are not limited to, customization based on user settings and editing the alert message. For example, the alert unit can customize the content of the alert using customization based on user settings and notify the user in an optimal manner. The alert unit can also use editing of the alert message to suggest optimal alert content based on the user's past reactions. The alert unit can also adjust the content of the alert in real time according to the user's situation. For example, the alert unit can customize the content of the alert using customization based on user settings and notify the user in an optimal manner. The alert message can be edited to suggest optimal alert content based on the user's past reactions. The alert content is adjusted in real time according to the user's situation. In this way, the alert content can be customized to notify the user in an optimal manner. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or may be performed without using AI. For example, the alert unit can customize the content of the alert using an AI model that uses user setting data as input and customizes the content of the alert.

[0050] When issuing an alert, the alert unit can select the optimal alert method by taking into account the user's geographical location information. Methods for considering geographical location information include, but are not limited to, the use of GPS data and location-based analysis algorithms. For example, the alert unit can use GPS data to prioritize a smartphone notification when the user is at home and issue an alert. The alert unit can also use a location-based analysis algorithm to prioritize a voice alert when the user is out. The alert unit can also combine the use of GPS data and a location-based analysis algorithm to suggest the optimal alert method based on the user's geographical location information. For example, the alert unit can use GPS data to prioritize a smartphone notification when the user is at home and issue an alert. The location-based analysis algorithm can also use a location-based analysis algorithm to prioritize a voice alert when the user is out. The GPS data and a location-based analysis algorithm can be combined to suggest the optimal alert method based on the user's geographical location information. This allows the optimal alert method to be selected by taking the user's geographical location information into account. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or without AI. For example, the alert unit can select an alert method using an AI model that inputs the user's geographic location data and selects the optimal alert method.

[0051] The alert unit can customize the alert method by reflecting the user's past feedback when issuing an alert. Methods for reflecting past feedback include, but are not limited to, storing and analyzing feedback data, updating user settings, and the like. For example, the alert unit can prioritize and use an alert method previously preferred by the user by storing and analyzing feedback data. The alert unit can also customize the alert method based on the user's past feedback by updating user settings. The alert unit can also learn the user's response patterns by combining storing and analyzing feedback data and updating user settings, and propose the optimal alert method. For example, the alert unit can prioritize and use an alert method previously preferred by the user by storing and analyzing feedback data, and customizing the alert method based on the user's past feedback by updating user settings. The alert unit can learn the user's response patterns by combining storing and analyzing feedback data and updating user settings, and propose the optimal alert method. This allows the optimal alert method to be selected by reflecting the user's past feedback. Some or all of the above-described processing in the alert unit may be performed, for example, using AI or without AI. For example, the alerting unit can customize the alerting means using an AI model that takes the user's past feedback data as input and customizes the alerting means.

[0052] The alert unit may be added with a function to simultaneously notify family members and related parties when an alert is issued. Methods for simultaneously notifying family members and related parties include, but are not limited to, group notification and multicast notification. For example, the alert unit may use group notification to notify family members by email when an alert occurs. The alert unit may also use multicast notification to send smartphone notifications to related parties when an alert occurs. The alert unit may also combine group notification and multicast notification to send audio warnings to family members and related parties when an alert occurs. For example, the alert unit may use group notification to notify family members by email when an alert occurs. The alert unit may use multicast notification to send smartphone notifications to related parties when an alert occurs. The alert unit may combine group notification and multicast notification to send audio warnings to family members and related parties when an alert occurs. This allows for simultaneous notification of family members and related parties, enabling a rapid response. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or may be performed without AI. For example, the alert unit may issue an alert using an AI model that takes the content of the alert as input and simultaneously notifies family members and related parties.

[0053] When intervening, the intervention unit can have the AI ​​automatically analyze the content of the conversation and intervene at the appropriate timing. Methods of intervening at the appropriate timing include, but are not limited to, timing adjustment based on conversation content analysis and intervention based on specific cues, for example. For example, the intervention unit can have the AI ​​automatically analyze the content of the conversation using timing adjustment based on conversation content analysis and intervene at the appropriate timing. Alternatively, the intervention unit can have the AI ​​automatically analyze the content of the conversation using intervention based on specific cues and intervene at the appropriate timing. Alternatively, the intervention unit can combine timing adjustment based on conversation content analysis and intervention based on specific cues to intervene at the appropriate timing. For example, the intervention unit can automatically analyze the content of the conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. Alternatively, the AI ​​can automatically analyze the content of the conversation using intervention based on specific cues and intervene at the appropriate timing. Alternatively, the AI ​​can combine timing adjustment based on conversation content analysis and intervention based on specific cues to automatically analyze the content of the conversation and intervene at the optimal timing. This allows the AI ​​to automatically analyze the content of the conversation and intervene at the appropriate time, thereby reducing the risk of fraud. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without using AI. For example, the intervention unit may use an AI model that inputs conversation content data and intervenes at an appropriate timing to adjust the timing of intervention.

[0054] At the time of intervention, the intervention unit can select an optimal intervention method by referring to past intervention history. Methods of referring to past intervention history include, but are not limited to, storing and searching intervention logs and analyzing historical data. For example, the intervention unit analyzes past intervention history by storing and searching intervention logs and selecting the most effective intervention method. The intervention unit can also prioritize the use of intervention methods to which the user has responded in the past by analyzing historical data. The intervention unit can also learn the user's response pattern by combining storing and searching intervention logs and analyzing historical data and propose the optimal intervention method. For example, the intervention unit analyzes past intervention history by storing and searching intervention logs and selecting the most effective intervention method. The intervention unit prioritizes the use of intervention methods to which the user has responded in the past by analyzing historical data. The intervention unit learns the user's response pattern by combining storing and searching intervention logs and analyzing historical data and proposes the optimal intervention method. In this way, the optimal intervention method can be selected by referring to past intervention history. Some or all of the above-described processing in the intervention unit may be performed, for example, using AI or without AI. For example, the intervention unit can select an intervention method using an AI model that inputs past intervention history data and selects the optimal intervention method.

[0055] The intervention unit may add a function that allows family members or related parties to remotely intervene during intervention. Examples of functions that allow remote intervention include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The intervention unit may also use a voice call to allow related parties to remotely intervene and support the elderly. The intervention unit may also configure the system to allow family members or related parties to remotely intervene using chat. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The related parties may use a voice call to remotely intervene and support the elderly. The system may be configured to allow family members or related parties to remotely intervene using chat. This allows family members or related parties to remotely intervene, enabling a rapid response. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may allow family members or related parties to remotely intervene using an AI model that sets up a video call, voice call, or chat for remote intervention.

[0056] When intervening, the intervention unit can automatically analyze the tone and speed of a conversation using AI and intervene at an appropriate time. Methods for analyzing the tone and speed of a conversation include, but are not limited to, extraction of speech features and tone analysis algorithms. For example, the intervention unit can analyze the tone of a conversation in real time using extraction of speech features and intervene when there is a high possibility of fraud. The intervention unit can also analyze the tone of a conversation in real time using a tone analysis algorithm and intervene at an appropriate time. The intervention unit can also analyze the speed of a conversation in real time and intervene at an appropriate time. For example, the intervention unit can analyze the tone of a conversation in real time using extraction of speech features and intervene when there is a high possibility of fraud. The intervention unit can analyze the tone of a conversation in real time using a tone analysis algorithm and intervene at an appropriate time. The intervention unit can analyze the speed of a conversation in real time and intervene at an appropriate time. In this way, the AI ​​can automatically analyze the tone and speed of a conversation and intervene at an appropriate time, thereby reducing the risk of fraud. Some or all of the above-described processing in the intervention unit may be performed using AI, for example, or without AI. For example, the intervention unit can adjust the timing of intervention using an AI model that takes data on the tone and speed of conversation as input and intervenes at the appropriate time.

[0057] The intervention unit may add a function that allows family members and related parties to intervene remotely. Examples of functions that allow remote intervention include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The intervention unit may also use a voice call to allow related parties to remotely intervene and support the elderly. The intervention unit may also configure the system to allow family members and related parties to remotely intervene using chat. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The related parties may use a voice call to remotely intervene and support the elderly. The system may be configured to allow family members and related parties to remotely intervene using chat. This allows family members and related parties to intervene remotely, enabling a rapid response. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may allow family members and related parties to remotely intervene using an AI model that sets up a video call, voice call, or chat for remote intervention.

[0058] The intervention unit allows the AI ​​to automatically analyze the content of a conversation and intervene at an appropriate time. Methods of intervening at an appropriate time include, but are not limited to, timing adjustment based on conversation content analysis and intervening based on a specific cue. For example, the intervention unit allows the AI ​​to automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene at an appropriate time. The intervention unit can also allow the AI ​​to automatically analyze the content of a conversation using intervention based on a specific cue and intervene at an appropriate time. The intervention unit can also intervene at an appropriate time by combining timing adjustment based on conversation content analysis and intervention based on a specific cue. For example, the intervention unit allows the AI ​​to automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. The AI ​​can automatically analyze the content of a conversation using intervention based on a specific cue and intervene at an appropriate time. The AI ​​can automatically analyze the content of a conversation and intervene at an optimal time by combining timing adjustment based on conversation content analysis and intervention based on a specific cue. This allows the AI ​​to automatically analyze the content of a conversation and intervene at an appropriate time, thereby reducing the risk of fraud. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without using AI. For example, the intervention unit may use an AI model that inputs conversation content data and intervenes at an appropriate timing to adjust the timing of intervention.

[0059] The system according to the embodiment is not limited to the above-described example, and various modifications are possible, for example, as follows.

[0060] During speech recognition, the speech recognition unit can learn the characteristics of a speaker's voice and use an individually optimized speech recognition model. For example, it can use acoustic feature extraction to learn the tone and pitch of a speaker's voice and generate an individually optimized speech recognition model. It can also use machine learning model training to learn the pronunciation habits and speed of a speaker and use an individually optimized speech recognition model. Furthermore, it can continuously learn the characteristics of a speaker's voice and update the speech recognition model. As a result, by learning the characteristics of a speaker's voice and using an individually optimized speech recognition model, the accuracy of speech recognition is improved.

[0061] The intervention unit allows the AI ​​to automatically analyze the content of a conversation and intervene at the appropriate time. For example, the AI ​​can automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. The AI ​​can also automatically analyze the content of a conversation and intervene at the appropriate time using intervention based on specific cues. Furthermore, timing adjustment based on conversation content analysis and intervention based on specific cues can be combined to intervene at the optimal time. This allows the AI ​​to automatically analyze the content of a conversation and intervene at the appropriate time, thereby reducing the risk of fraud.

[0062] The speech recognition unit can add a filtering function that automatically removes background noise during speech recognition. For example, a noise canceling algorithm can be used to automatically remove background noise during speech recognition. It can also use filtering technology to remove noise in specific frequency bands during speech recognition. Furthermore, it can learn environmental sounds and automatically remove specific patterns of noise. This automatically removes background noise, improving the accuracy of speech recognition.

[0063] During the analysis, the analysis unit can analyze the tone and speed of the conversation to determine the possibility of fraud. For example, speech feature extraction can be used to analyze the tone of the conversation to determine the possibility of fraud. A tone analysis algorithm can also be used to analyze the tone of the conversation in real time to determine the possibility of fraud. Furthermore, the speed of the conversation can also be analyzed in real time to determine the possibility of fraud. Thus, by analyzing the tone and speed of the conversation, the possibility of fraud can be more accurately determined.

[0064] The alert unit can combine multiple notification methods to issue an alert when an alert occurs. For example, a smartphone notification and email can be sent simultaneously, allowing the user to be notified quickly. It can also combine a voice alert with a smartphone notification to notify the user both visually and audibly. It can also combine an email and a voice alert to reliably notify the user of the alert. In this way, by combining multiple notification methods, it is possible to quickly and reliably notify the user of the alert.

[0065] During analysis, the analysis unit can improve analysis accuracy by learning not only past fraud cases but also the latest fraud techniques. For example, it can improve analysis accuracy by collecting and learning from the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques. It can also improve analysis accuracy by using an online learning algorithm to learn the latest fraud techniques in real time. Furthermore, it can also update the analysis model by combining past fraud cases and the latest fraud techniques. This improves analysis accuracy by learning the latest fraud techniques.

[0066] The processing flow of the first embodiment will be briefly explained below.

[0067] Step 1: The speech recognition unit recognizes the voice conversation in real time, for example, by using a deep learning model to convert speech to text, and converts what the elderly person is saying on the phone into text in real time. Step 2: The analysis unit analyzes the speech recognized by the speech recognition unit. For example, it uses a model trained on past fraud cases to determine the possibility of fraud, and if the speech contains phrases such as "Please transfer money" or "Please respond quickly," it determines that there is a high possibility of fraud. Step 3: The alert unit issues an alert if a fraud risk is detected based on the results of the analysis by the analysis unit. For example, it can send a notification to the elderly person's smartphone, send an email, or send a voice warning, or send an email to their family. Step 4: The intervention unit intervenes in the conversation when an alert is issued by the alert unit. For example, the AI ​​can automatically interrupt the conversation and warn, "This conversation may be fraudulent." Family members can also remotely join the conversation and provide support to the elderly.

[0068] (Example 2) A fraud prevention system according to an embodiment of the present invention is a system for preventing "it's my son" frauds committed by elderly people. This fraud prevention system recognizes voice conversations in real time, analyzes their content using AI, issues an alert if there is a possibility of fraud, and intervenes in the conversation as necessary. For example, the fraud prevention system recognizes voice conversations in real time. For example, the fraud prevention system converts what an elderly person is saying on the phone into text in real time, and then analyzes the text using AI. The fraud prevention system then determines whether there is a possibility of fraud based on the text analyzed by the AI. For example, if the text contains phrases such as "please transfer money" or "please handle this quickly," the fraud prevention system determines that there is a high possibility of fraud. Next, if the fraud prevention system determines that there is a possibility of fraud, it issues an alert. For example, it sends a notification to the elderly person's smartphone or notifies their family by email. Furthermore, the fraud prevention system intervenes in the conversation as necessary. For example, the AI ​​automatically interrupts the conversation and warns them that "this conversation may be fraudulent." Family members can also remotely participate in the conversation and provide support to the elderly. In this way, the fraud prevention system can prevent "it's my son" frauds committed by elderly people. This allows the fraud prevention system to prevent "it's me" scams against the elderly. For example, if a fraudulent phone call is received, the fraud prevention system will automatically respond, preventing the elderly from becoming victims. The elderly can then live their daily lives with peace of mind, and their families can watch over them with peace of mind.

[0069] The fraud prevention system according to the embodiment includes a voice recognition unit, an analysis unit, an alert unit, and an intervention unit. The voice recognition unit recognizes voice conversations in real time. The voice recognition unit converts voice into text, for example, using a deep learning model. The voice recognition unit converts, for example, what an elderly person is saying on the phone into text in real time. The voice recognition unit converts voice into text with high accuracy, for example, using a deep learning model. The analysis unit analyzes the voice recognized by the voice recognition unit. The analysis unit determines the possibility of fraud, for example, using a model trained on past fraud cases. The analysis unit determines the possibility of fraud if the voice contains phrases such as "please transfer money" or "please handle this quickly." The analysis unit determines the possibility of fraud with high accuracy, for example, using a model trained on past fraud cases. The alert unit issues an alert if a risk of fraud is detected based on the results of the analysis by the analysis unit. The alert unit issues a warning, for example, via smartphone notification, email, or voice alert. The alert unit, for example, sends a notification to the elderly person's smartphone or notifies family members by email. The alert unit can quickly notify the user of the possibility of fraud by, for example, issuing a smartphone notification, email, or voice warning. The intervention unit intervenes in the conversation when the alert unit issues an alert. For example, the intervention unit automatically interrupts the conversation using AI to warn the user that "this conversation may be fraudulent." For example, the intervention unit allows a family member to remotely participate in the conversation and support the elderly. For example, the intervention unit automatically interrupts the conversation using AI to warn the user of the possibility of fraud, thereby reducing the risk of elderly people falling victim to fraud. As a result, the fraud prevention system according to the embodiment can prevent "it's my son" frauds committed by elderly people.

[0070] The speech recognition unit can convert speech into text using a deep learning model. Examples of deep learning models include, but are not limited to, a convolutional neural network (CNN), a recurrent neural network (RNN), and a transformer. The speech recognition unit can, for example, use a CNN to analyze speech data and convert it into text. The speech recognition unit can also use an RNN to analyze temporal dependencies in speech data and convert it into text. The speech recognition unit can also use a Transformer to analyze speech data with high accuracy and convert it into text. For example, the speech recognition unit can use a CNN to extract features of speech data and convert it into text. The RNN analyzes temporal dependencies in speech data and converts continuous speech data into text. The Transformer analyzes speech data with high accuracy and converts it into text taking context into account. In this way, the deep learning model can convert speech into text with high accuracy. Some or all of the above-mentioned processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can convert speech to text using an AI model that uses a deep learning model to input speech data and outputs text data.

[0071] The analysis unit can determine the likelihood of fraud using a model trained on past fraud cases. Examples of models trained on past fraud cases include, but are not limited to, fraud case datasets and supervised learning. The analysis unit, for example, trains a model using a fraud case dataset to determine the likelihood of fraud. The analysis unit can also use supervised learning to learn fraud patterns and determine the likelihood of fraud. The analysis unit can also combine a fraud case dataset and supervised learning to determine the likelihood of fraud with high accuracy. For example, the analysis unit trains a model using a fraud case dataset to determine the likelihood of fraud when specific keywords or phrases are included. Supervised learning learns fraud patterns and enables adaptation to new fraud methods. Combining a fraud case dataset and supervised learning can determine the likelihood of fraud with high accuracy. As a result, using a model trained on past fraud cases can determine the likelihood of fraud with high accuracy. Some or all of the above-described processing in the analysis unit may be performed, for example, using AI or without AI. For example, the analysis unit can determine the likelihood of fraud using an AI model that takes a fraud case dataset as input and outputs the likelihood of fraud.

[0072] The alert unit can issue a smartphone notification, email, or voice alert. Examples of smartphone notifications include, but are not limited to, push notifications, SMS, and in-app notifications. The alert unit can send an alert to the elderly person's smartphone using, for example, push notifications. The alert unit can also send an alert to the elderly person's smartphone using SMS. The alert unit can also send an alert to the elderly person's smartphone using an in-app notification. Examples of emails include, but are not limited to, HTML emails and text emails. The alert unit can send an alert to family members using, for example, HTML emails. The alert unit can also send an alert to family members using text emails. Examples of voice alerts include, but are not limited to, voice synthesis and recorded voices. The alert unit can send an alert to the elderly person using, for example, voice synthesis. The alert unit can also send an alert to the elderly person using recorded voices. This allows the user to be quickly notified of a possible fraud by issuing a smartphone notification, email, or voice alert. Some or all of the above-described processing by the alert unit can be performed using, for example, AI, or without AI. For example, the alerting unit can use AI models to issue alerts via smartphone notifications, emails, and voice warnings when potential fraud is detected.

[0073] The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud." Methods for the AI ​​to automatically interrupt the conversation include, but are not limited to, when a specific keyword appears or after a certain period of time has passed. For example, the intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a specific keyword appears. The intervention unit can also automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," after a certain period of time has passed. The intervention unit can also automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when the AI ​​analyzes the content of the conversation and determines that there is a high risk of fraud. For example, the intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a specific keyword appears. The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when a certain period of time has passed. The intervention unit can automatically interrupt the conversation and warn, "This conversation poses a risk of fraud," when the AI ​​analyzes the content of the conversation and determines that there is a high risk of fraud. In this way, the AI ​​can automatically interrupt the conversation and warn, reducing the risk of elderly people falling victim to fraud. Some or all of the above-described processing by the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may interrupt a conversation using an AI model that takes the content of the conversation as input and outputs a warning about the risk of fraud.

[0074] The intervention unit can support the elderly by having a family member remotely participate in the conversation. Methods for remotely participating in the conversation include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using a video call. The intervention unit can also support the elderly by having a family member remotely participate in the conversation using a voice call. The intervention unit can also support the elderly by having a family member remotely participate in the conversation using a chat. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using a video call and warn of the risk of fraud. The family can support the elderly by having a family member remotely participate in the conversation using a voice call. The family can support the elderly by having a family member remotely participate in the conversation using a chat. In this way, the elderly can be supported by having a family member remotely participate in the conversation and the risk of fraud can be further reduced. Some or all of the above-described processing in the intervention unit can be performed using, for example, AI, or without AI. For example, the intervention unit can support the elderly by having a family member remotely participate in the conversation using an AI model that sets up a video call, voice call, or chat for remotely participating in the conversation.

[0075] The voice recognition unit can estimate the user's emotion and adjust the accuracy of voice recognition based on the estimated user's emotion. Methods for estimating the user's emotion include, but are not limited to, voice analysis, facial expression recognition, and text analysis. The voice recognition unit can, for example, use voice analysis to estimate the user's emotion and adjust the accuracy of voice recognition. The voice recognition unit can also use facial expression recognition to estimate the user's emotion and adjust the accuracy of voice recognition. The voice recognition unit can also use text analysis to estimate the user's emotion and adjust the accuracy of voice recognition. For example, if the user is nervous using voice analysis, the voice recognition unit can increase the sensitivity of voice recognition to more accurately recognize voice. If the user is relaxed using facial expression recognition, the voice recognition unit can return the sensitivity to normal and recognize natural conversation. If the user is excited using text analysis, the voice recognition unit can adjust the sensitivity of voice recognition to reduce erroneous recognition. As a result, adjusting the accuracy of voice recognition based on the user's emotion enables more accurate voice recognition. Some or all of the above-described processing in the voice recognition unit may be performed using, for example, AI or without AI. For example, the voice recognition unit can adjust the accuracy of voice recognition using an AI model that takes user emotional data as input and adjusts the accuracy of voice recognition.

[0076] The voice recognition unit may add a filtering function that automatically removes background noise during voice recognition. Methods for automatically removing background noise include, but are not limited to, noise canceling algorithms and filtering techniques. The voice recognition unit may, for example, use a noise canceling algorithm to automatically remove background noise during voice recognition. The voice recognition unit may also use filtering techniques to remove noise in a specific frequency band during voice recognition. The voice recognition unit may also learn environmental sounds and automatically remove noise of a specific pattern. For example, the voice recognition unit may use a noise canceling algorithm to analyze ambient noise in real time and remove noise. Filtering techniques may be used to remove noise in a specific frequency band and recognize only speech. Environmental sounds may be learned and noise of a specific pattern may be automatically removed. This automatically removing background noise improves the accuracy of voice recognition. Some or all of the above-described processing in the voice recognition unit may be performed using, for example, AI, or may be performed without AI. For example, the voice recognition unit may automatically remove background noise using an AI model that receives background noise as input and outputs noise-removed voice data.

[0077] During speech recognition, the speech recognition unit can learn the characteristics of a speaker's voice and use an individually optimized speech recognition model. Methods for learning the characteristics of a speaker's voice include, but are not limited to, extraction of acoustic features and training of a machine learning model. For example, the speech recognition unit can learn the tone and pitch of a speaker's voice using the extraction of acoustic features and generate an individually optimized speech recognition model. The speech recognition unit can also learn the pronunciation habits and speed of a speaker using training of a machine learning model and use an individually optimized speech recognition model. The speech recognition unit can also continuously learn the characteristics of a speaker's voice and update the speech recognition model. For example, the speech recognition unit can learn the tone and pitch of a speaker using the extraction of acoustic features and generate an individually optimized speech recognition model. The speech recognition unit can also learn the pronunciation habits and speed of a speaker using training of a machine learning model and use an individually optimized speech recognition model. The speech recognition unit continuously learns the characteristics of a speaker's voice and updates the speech recognition model. As a result, by learning the characteristics of a speaker's voice and using an individually optimized speech recognition model, the accuracy of speech recognition is improved. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that receives input of feature data of a speaker's voice and outputs an individually optimized speech recognition model.

[0078] The speech recognition unit may apply an algorithm that takes into account the context of a conversation during speech recognition to reduce misrecognition. Methods for taking into account the context of a conversation include, but are not limited to, analysis of preceding and following utterances and context-based algorithms. For example, the speech recognition unit may apply an algorithm that takes into account the context of a conversation and reduces misrecognition. The speech recognition unit may also use a context-based algorithm to analyze the content of a conversation in real time and improve the accuracy of speech recognition based on the context. The speech recognition unit may also learn the topic of a conversation and reduce misrecognition based on a specific context. For example, the speech recognition unit may use an analysis of preceding and following utterances to apply an algorithm that takes into account the context of a conversation and reduces misrecognition. The context-based algorithm may analyze the content of a conversation in real time and improve the accuracy of speech recognition based on the context. The topic of a conversation may be learned and misrecognition may be reduced based on a specific context. Thus, misrecognition can be reduced by taking the context of a conversation into account. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI or without AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that takes conversational context data as input and outputs an algorithm that reduces misrecognition.

[0079] The voice recognition unit can estimate a user's emotion and adjust the timing of voice recognition based on the estimated user's emotion. Methods for adjusting the timing of voice recognition include, but are not limited to, timing adjustment according to changes in emotion and adjustment based on specific cues. For example, the voice recognition unit estimates a user's emotion using timing adjustment according to changes in emotion and adjusts the timing of voice recognition. The voice recognition unit can also estimate a user's emotion using adjustment based on specific cues and adjust the timing of voice recognition. The voice recognition unit can also adjust the timing of voice recognition by combining timing adjustment according to changes in emotion and adjustment based on specific cues. For example, if the user is nervous, the voice recognition unit can advance the timing of voice recognition to respond quickly. If the user is relaxed, the voice recognition unit can return the timing of voice recognition to normal and maintain natural conversation. If the user is excited, the voice recognition unit can combine timing adjustment according to changes in emotion and adjustment based on specific cues to adjust the timing of voice recognition to reduce erroneous recognition. This enables more accurate voice recognition by adjusting the timing of voice recognition based on the user's emotion. Some or all of the above-described processing in the voice recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the voice recognition unit may adjust the timing of voice recognition using an AI model that receives user emotion data as input and adjusts the timing of voice recognition.

[0080] The speech recognition unit can improve the recognition accuracy by taking into account the speaker's geographical accent or dialect during speech recognition. Methods for taking into account geographical accents or dialects include, but are not limited to, learning regional speech data and using accent conversion algorithms. For example, the speech recognition unit learns the speaker's geographical accent using regional speech data learning and reflects it in a speech recognition model. The speech recognition unit can also analyze the speaker's dialect using an accent conversion algorithm to customize the speech recognition model. The speech recognition unit can also apply an algorithm that improves the speech recognition accuracy based on the geographical accent or dialect. For example, the speech recognition unit learns regional speech data, learns the speaker's geographical accent, and reflects it in a speech recognition model. The speech recognition unit analyzes the speaker's dialect using an accent conversion algorithm and customizes the speech recognition model. An algorithm that improves the speech recognition accuracy based on the geographical accent or dialect is applied. As a result, the accuracy of speech recognition is improved by taking into account the speaker's geographical accent or dialect. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or without AI. For example, the speech recognition unit can use AI models that take geographic accent and dialect data as input and customize the speech recognition model to improve speech recognition accuracy.

[0081] The speech recognition unit can customize a speech recognition model based on the age and gender of the speaker during speech recognition. Methods of customization based on age and gender include, but are not limited to, speech models for each age group and speech features for each gender. For example, the speech recognition unit customizes the speech recognition model based on the speaker's age using a speech model for each age group. The speech recognition unit can also optimize the speech recognition model based on the speaker's gender using speech features for each gender. The speech recognition unit can also apply an algorithm that improves speech recognition accuracy by taking age and gender into consideration. For example, the speech recognition unit customizes the speech recognition model based on the speaker's age using a speech model for each age group. The speech recognition unit optimizes the speech recognition model based on the speaker's gender using speech features for each gender. The speech recognition unit applies an algorithm that improves speech recognition accuracy by taking age and gender into consideration. Customizing the speech recognition model based on the speaker's age and gender improves speech recognition accuracy. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit can improve the accuracy of speech recognition by using an AI model that takes data such as the speaker's age and gender as input and customizes the speech recognition model.

[0082] During speech recognition, the speech recognition unit can improve recognition accuracy by referring to a speaker's past conversation history. Methods for referring to the past conversation history include, but are not limited to, storing and searching a conversation log and analyzing historical data. For example, the speech recognition unit learns the speaker's past conversation history using the storing and searching of a conversation log and reflects it in a speech recognition model. The speech recognition unit can also analyze the content of the speaker's past conversations using analysis of historical data to improve speech recognition accuracy. The speech recognition unit can also combine the storing and searching of a conversation log and the analysis of historical data to learn the speaker's past conversation patterns and apply an algorithm that reduces recognition errors. For example, the speech recognition unit learns the speaker's past conversation history using the storing and searching of a conversation log and reflects it in a speech recognition model. The content of the speaker's past conversations using analysis of historical data can be analyzed to improve speech recognition accuracy. By combining the storing and searching of a conversation log and the analysis of historical data to learn the speaker's past conversation patterns and apply an algorithm that reduces recognition errors, speech recognition accuracy is improved. In this way, the speech recognition accuracy is improved by referring to the speaker's past conversation history. Some or all of the above-described processing in the speech recognition unit may be performed using, for example, AI, or may be performed without using AI. For example, the speech recognition unit may use an AI model that uses the speaker's past conversation history data as input and customizes the speech recognition model to improve the accuracy of speech recognition.

[0083] The analysis unit can estimate the user's emotion and adjust the reliability of the analysis result based on the estimated user's emotion. Methods for adjusting the reliability of the analysis result include, but are not limited to, reliability adjustment based on an emotion score and parameter adjustment of an analysis algorithm. For example, the analysis unit estimates the user's emotion using reliability adjustment based on an emotion score and adjusts the reliability of the analysis result. The analysis unit can also estimate the user's emotion using parameter adjustment of an analysis algorithm and adjust the reliability of the analysis result. The analysis unit can also adjust the reliability of the analysis result by combining reliability adjustment based on an emotion score and parameter adjustment of an analysis algorithm. For example, the analysis unit can increase the reliability of the analysis result when the user is nervous by using reliability adjustment based on an emotion score to provide more accurate results. When the user is relaxed by using parameter adjustment of the analysis algorithm, the analysis unit can return the reliability of the analysis result to normal and perform a natural analysis. When the user is excited by combining reliability adjustment based on an emotion score and parameter adjustment of an analysis algorithm, the analysis unit can adjust the reliability of the analysis result to reduce erroneous determinations. This allows for more accurate analysis results to be provided by adjusting the reliability of the analysis results based on the user's emotions. Some or all of the above-described processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit may adjust the reliability of the analysis results using an AI model that uses the user's emotional data as input and adjusts the reliability of the analysis results.

[0084] During analysis, the analysis unit can apply an algorithm that takes into account the context of the conversation to more accurately determine the likelihood of fraud. Methods that take into account the context of the conversation include, but are not limited to, analysis of surrounding utterances and context-based algorithms. For example, the analysis unit applies an algorithm that takes into account the context of the conversation and more accurately determines the likelihood of fraud. The analysis unit can also use a context-based algorithm to analyze the content of the conversation in real time and determine the likelihood of fraud based on the context. The analysis unit can also learn the topic of the conversation and determine the likelihood of fraud based on the specific context. For example, the analysis unit applies an algorithm that takes into account the context of the conversation and more accurately determines the likelihood of fraud by using analysis of surrounding utterances. The context-based algorithm analyzes the content of the conversation in real time and determines the likelihood of fraud based on the context. The topic of the conversation is learned and the likelihood of fraud is determined based on the specific context. In this way, by taking the context of the conversation into account, the likelihood of fraud can be more accurately determined. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI or without AI. For example, the analysis unit can more accurately determine the likelihood of fraud by using an AI model that takes conversation context data as input and outputs an algorithm that determines the likelihood of fraud.

[0085] During analysis, the analysis unit can learn not only past fraud cases but also the latest fraud techniques to improve analysis accuracy. Methods for learning the latest fraud techniques include, but are not limited to, collecting and learning from the latest fraud datasets and online learning algorithms. For example, the analysis unit can use the collection and learning of the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques, thereby improving analysis accuracy. The analysis unit can also use an online learning algorithm to learn the latest fraud techniques in real time to improve analysis accuracy. The analysis unit can also combine past fraud cases with the latest fraud techniques to update the analysis model. For example, the analysis unit can use the collection and learning of the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques, thereby improving analysis accuracy. The online learning algorithm can learn the latest fraud techniques in real time to improve analysis accuracy. The analysis model is updated by combining past fraud cases with the latest fraud techniques. In this way, by learning the latest fraud techniques, analysis accuracy is improved. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can improve analysis accuracy by using an AI model that takes the latest fraud data set as input and updates the analysis model.

[0086] During analysis, the analysis unit can analyze the tone and speed of speech to determine the possibility of fraud. Methods for analyzing the tone and speed of speech include, but are not limited to, extraction of speech features and tone analysis algorithms. For example, the analysis unit can analyze the tone of speech using extraction of speech features to determine the possibility of fraud. The analysis unit can also analyze the tone of speech in real time using a tone analysis algorithm to determine the possibility of fraud. The analysis unit can also analyze the speed of speech in real time to determine the possibility of fraud. For example, the analysis unit can analyze the tone of speech using extraction of speech features to determine the possibility of fraud. The analysis unit can analyze the tone of speech in real time using a tone analysis algorithm to determine the possibility of fraud. The analysis unit can analyze the speed of speech in real time to determine the possibility of fraud. In this way, by analyzing the tone and speed of speech, the possibility of fraud can be more accurately determined. Some or all of the above-described processing in the analysis unit may be performed using, for example, AI, or may be performed without AI. For example, the analysis unit can input data on the tone and speed of speech and use an AI model that determines the possibility of fraud to determine the possibility of fraud.

[0087] The analysis unit can estimate the user's emotion and adjust the display method of the analysis results based on the estimated user's emotion. Methods for adjusting the display method of the analysis results include, but are not limited to, display adjustment based on an emotion score and user interface customization. For example, the analysis unit can estimate the user's emotion using display adjustment based on an emotion score and adjust the display method of the analysis results. The analysis unit can also estimate the user's emotion using user interface customization and adjust the display method of the analysis results. The analysis unit can also adjust the display method of the analysis results by combining display adjustment based on an emotion score and user interface customization. For example, the analysis unit can use display adjustment based on an emotion score to provide a simple, highly visible display method when the user is nervous. Use user interface customization to provide a display method containing detailed information when the user is relaxed. Combine display adjustment based on an emotion score and user interface customization to provide a visually stimulating display method when the user is excited. This allows for more appropriate display by adjusting the display method of the analysis results based on the user's emotion. Some or all of the above-described processing in the analysis unit can be performed using, for example, AI, or without AI. For example, the analysis unit can adjust the way the analysis results are displayed using an AI model that takes the user's emotional data as input and adjusts the way the analysis results are displayed.

[0088] During analysis, the analysis unit can determine the possibility of fraud by analyzing not only the content of the conversation but also the speaker's tone of voice and emotion. Methods for analyzing tone of voice and emotion include, but are not limited to, extraction of voice features and emotion analysis algorithms. For example, the analysis unit can analyze the content of the conversation using extraction of voice features and determine the possibility of fraud by taking into account the speaker's tone of voice. The analysis unit can also use an emotion analysis algorithm to analyze the speaker's emotion in real time and determine the possibility of fraud. The analysis unit can also combine the content of the conversation and the speaker's emotion to more accurately determine the possibility of fraud. For example, the analysis unit can analyze the content of the conversation using extraction of voice features and determine the possibility of fraud by taking into account the speaker's tone of voice. The emotion analysis algorithm can analyze the speaker's emotion in real time and determine the possibility of fraud. The combination of the content of the conversation and the speaker's emotion can more accurately determine the possibility of fraud. In this way, by analyzing the speaker's tone of voice and emotion, the possibility of fraud can be more accurately determined. Some or all of the above-mentioned processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can input data on the content of the conversation, tone of voice, and emotions, and use an AI model that determines the possibility of fraud to determine the possibility of fraud.

[0089] During analysis, the analysis unit can determine the possibility of fraud by referring to the speaker's past behavioral history. Methods of referring to past behavioral history include, but are not limited to, storing and searching behavior logs and analyzing history data. For example, the analysis unit can learn the speaker's past behavioral history using the storing and searching of behavior logs and determine the possibility of fraud. The analysis unit can also analyze the speaker's past behavioral patterns using the analysis of history data to predict the possibility of fraud. The analysis unit can also combine the storing and searching of behavior logs and the analysis of history data to refer to the speaker's past behavioral history in real time and determine the possibility of fraud. For example, the analysis unit can use the storing and searching of behavior logs and the analysis of history data to learn the speaker's past behavioral history and determine the possibility of fraud. The analysis unit can use the storing and searching of behavior logs and the analysis of history data to analyze the speaker's past behavioral patterns and predict the possibility of fraud. The analysis unit can combine the storing and searching of behavior logs and the analysis of history data to refer to the speaker's past behavioral history in real time and determine the possibility of fraud. By doing so, the possibility of fraud can be more accurately determined by referring to the speaker's past behavioral history. Some or all of the above-described processing in the analysis unit may be performed using, for example, AI, or may be performed without using AI. For example, the analysis unit can input the speaker's past behavioral history data and determine the possibility of fraud using an AI model that determines the possibility of fraud.

[0090] During analysis, the analysis unit can determine the possibility of fraud by taking into account the speaker's geographic location information. Methods for taking into account geographic location information include, but are not limited to, the use of GPS data and a location-based analysis algorithm. For example, the analysis unit can use GPS data to analyze the speaker's geographic location information and determine the possibility of fraud. The analysis unit can also use a location-based analysis algorithm to refer to the speaker's location information in real time and predict the possibility of fraud. The analysis unit can also combine the use of GPS data and a location-based analysis algorithm to determine the possibility of fraud by taking into account the speaker's geographic location information. For example, the analysis unit can use GPS data to analyze the speaker's geographic location information and determine the possibility of fraud. The location-based analysis algorithm can refer to the speaker's location information in real time and predict the possibility of fraud. The use of GPS data and a location-based analysis algorithm can be combined to determine the possibility of fraud by taking into account the speaker's geographic location information. This allows for a more accurate determination of the possibility of fraud by taking into account the speaker's geographic location information. Some or all of the above-described processing in the analysis unit can be performed, for example, using AI or without AI. For example, the analysis unit can use an AI model that takes the speaker's geographic location data as input and determines the likelihood of fraud.

[0091] The alert unit can estimate the user's emotion and adjust the intensity of the alert based on the estimated user's emotion. Methods for adjusting the intensity of the alert include, but are not limited to, intensity adjustment based on an emotion score and adjustment of the alert volume. For example, the alert unit can estimate the user's emotion and adjust the intensity of the alert using intensity adjustment based on an emotion score. The alert unit can also estimate the user's emotion and adjust the intensity of the alert using adjustment of the alert volume. The alert unit can also adjust the intensity of the alert by combining intensity adjustment based on an emotion score and adjustment of the alert volume. For example, the alert unit can increase the intensity of the alert when the user is nervous by using intensity adjustment based on the emotion score to quickly attract attention. The alert unit can return the intensity of the alert to normal when the user is relaxed by using adjustment of the alert volume. The alert unit can adjust the intensity of the alert when the user is excited by combining intensity adjustment based on the emotion score and adjustment of the alert volume to avoid an overreaction. This allows for more appropriate alerts by adjusting the intensity of the alert based on the user's emotion. Some or all of the above-described processing in the alert unit can be performed using, for example, AI, or without AI. For example, the alert unit can adjust the intensity of the alert using an AI model that takes user emotion data as input and adjusts the intensity of the alert.

[0092] When issuing an alert, the alert unit can combine multiple notification means to issue the alert. Examples of multiple notification means include, but are not limited to, smartphone notifications, emails, and audio alerts. For example, the alert unit can simultaneously send a smartphone notification and an email to quickly notify the user of the alert. The alert unit can also combine an audio alert and a smartphone notification to quickly notify the user of the alert both visually and audibly. The alert unit can also combine an email and an audio alert to reliably notify the user of the alert. For example, the alert unit can simultaneously send a smartphone notification and an email to quickly notify the user of the alert. The audio alert and a smartphone notification can be combined to quickly notify the user of the alert both visually and audibly. The email and audio alert can be combined to reliably notify the user of the alert. In this way, by combining multiple notification means, the alert can be quickly and reliably notified of the alert. Some or all of the above-described processing in the alert unit may be performed, for example, using AI, or may be performed without AI. For example, the alert unit can issue an alert using an AI model that receives the alert content as input and outputs an alert using multiple notification means.

[0093] When issuing an alert, the alert unit can select the most appropriate alert method by referring to past alert history. Methods for referencing past alert history include, but are not limited to, saving and searching alert logs and analyzing historical data. The alert unit, for example, analyzes past alert history by saving and searching alert logs and selects the most effective notification method. The alert unit can also prioritize alert methods to which the user has responded in the past by analyzing historical data. The alert unit can also learn user response patterns by combining saving and searching alert logs and analyzing historical data and propose the most appropriate alert method. For example, the alert unit analyzes past alert history by saving and searching alert logs and selects the most effective notification method. Prioritizes alert methods to which the user has responded in the past by analyzing historical data. Learns user response patterns by combining saving and searching alert logs and analyzing historical data and proposes the most appropriate alert method. In this way, the most appropriate alert method can be selected by referring to past alert history. Some or all of the above-described processing in the alert unit may be performed, for example, using AI or without AI. For example, the alert unit can select an alert method using an AI model that inputs past alert history data and selects the optimal alert method.

[0094] The alert unit can customize the content of the alert when issuing an alert and notify the user in an optimal manner. Methods for customizing the content of the alert include, but are not limited to, customization based on user settings and editing the alert message. For example, the alert unit can customize the content of the alert using customization based on user settings and notify the user in an optimal manner. The alert unit can also use editing of the alert message to suggest optimal alert content based on the user's past reactions. The alert unit can also adjust the content of the alert in real time according to the user's situation. For example, the alert unit can customize the content of the alert using customization based on user settings and notify the user in an optimal manner. The alert message can be edited to suggest optimal alert content based on the user's past reactions. The alert content is adjusted in real time according to the user's situation. In this way, the alert content can be customized to notify the user in an optimal manner. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or may be performed without using AI. For example, the alert unit can customize the content of the alert using an AI model that uses user setting data as input and customizes the content of the alert.

[0095] The alert unit can estimate the user's emotion and adjust the timing of the alert based on the estimated user's emotion. Methods for adjusting the timing of the alert include, but are not limited to, timing adjustment based on an emotion score and adjustment based on a specific event. For example, the alert unit can estimate the user's emotion and adjust the timing of the alert using timing adjustment based on an emotion score. The alert unit can also estimate the user's emotion and adjust the timing of the alert using adjustment based on a specific event. The alert unit can also adjust the timing of the alert by combining timing adjustment based on an emotion score and adjustment based on a specific event. For example, if the user is nervous, the alert unit can advance the timing of the alert to quickly attract attention. If the user is relaxed, the alert unit can return the timing of the alert to normal and provide a natural notification. If the user is excited, the timing adjustment based on the emotion score and adjustment based on a specific event can be combined to adjust the timing of the alert to avoid an overreaction. This allows for more appropriate alerts by adjusting the timing of the alert based on the user's emotion. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or may be performed without AI. For example, the alert unit can adjust the timing of the alert using an AI model that takes user emotion data as input and adjusts the timing of the alert.

[0096] When issuing an alert, the alert unit can select the optimal alert method by taking into account the user's geographical location information. Methods for considering geographical location information include, but are not limited to, the use of GPS data and location-based analysis algorithms. For example, the alert unit can use GPS data to prioritize a smartphone notification when the user is at home and issue an alert. The alert unit can also use a location-based analysis algorithm to prioritize a voice alert when the user is out. The alert unit can also combine the use of GPS data and a location-based analysis algorithm to suggest the optimal alert method based on the user's geographical location information. For example, the alert unit can use GPS data to prioritize a smartphone notification when the user is at home and issue an alert. The location-based analysis algorithm can also use a location-based analysis algorithm to prioritize a voice alert when the user is out. The GPS data and a location-based analysis algorithm can be combined to suggest the optimal alert method based on the user's geographical location information. This allows the optimal alert method to be selected by taking the user's geographical location information into account. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or without AI. For example, the alert unit can select an alert method using an AI model that inputs the user's geographic location data and selects the optimal alert method.

[0097] The alert unit can customize the alert method by reflecting the user's past feedback when issuing an alert. Methods for reflecting past feedback include, but are not limited to, storing and analyzing feedback data, updating user settings, and the like. For example, the alert unit can prioritize and use an alert method previously preferred by the user by storing and analyzing feedback data. The alert unit can also customize the alert method based on the user's past feedback by updating user settings. The alert unit can also learn the user's response patterns by combining storing and analyzing feedback data and updating user settings, and propose the optimal alert method. For example, the alert unit can prioritize and use an alert method previously preferred by the user by storing and analyzing feedback data, and customizing the alert method based on the user's past feedback by updating user settings. The alert unit can learn the user's response patterns by combining storing and analyzing feedback data and updating user settings, and propose the optimal alert method. This allows the optimal alert method to be selected by reflecting the user's past feedback. Some or all of the above-described processing in the alert unit may be performed, for example, using AI or without AI. For example, the alerting unit can customize the alerting means using an AI model that takes the user's past feedback data as input and customizes the alerting means.

[0098] The alert unit may be added with a function to simultaneously notify family members and related parties when an alert is issued. Methods for simultaneously notifying family members and related parties include, but are not limited to, group notification and multicast notification. For example, the alert unit may use group notification to notify family members by email when an alert occurs. The alert unit may also use multicast notification to send smartphone notifications to related parties when an alert occurs. The alert unit may also combine group notification and multicast notification to send audio warnings to family members and related parties when an alert occurs. For example, the alert unit may use group notification to notify family members by email when an alert occurs. The alert unit may use multicast notification to send smartphone notifications to related parties when an alert occurs. The alert unit may combine group notification and multicast notification to send audio warnings to family members and related parties when an alert occurs. This allows for simultaneous notification of family members and related parties, enabling a rapid response. Some or all of the above-described processing in the alert unit may be performed using, for example, AI, or may be performed without AI. For example, the alert unit may issue an alert using an AI model that takes the content of the alert as input and simultaneously notifies family members and related parties.

[0099] The intervention unit can estimate the user's emotions and adjust the intervention method based on the estimated user's emotions. Methods for adjusting the intervention method include, but are not limited to, methods such as adjusting the method based on an emotion score and editing an intervention message. For example, the intervention unit can estimate the user's emotions and adjust the intervention method using method adjustment based on an emotion score. The intervention unit can also estimate the user's emotions and adjust the intervention method using editing the intervention message. The intervention unit can also adjust the intervention method by combining method adjustment based on an emotion score and editing the intervention message. For example, the intervention unit can use method adjustment based on an emotion score to intervene in a calm tone when the user is nervous. The intervention unit can use editing the intervention message to intervene in a normal tone when the user is relaxed. The intervention unit can combine method adjustment based on an emotion score and editing the intervention message to intervene in a calm tone when the user is excited. This allows for more appropriate intervention by adjusting the intervention method based on the user's emotions. Some or all of the above-described processing in the intervention unit can be performed using, for example, AI, or without AI. For example, the intervention unit can adjust the method of intervention using an AI model that takes the user's emotional data as input and adjusts the method of intervention.

[0100] When intervening, the intervention unit can have the AI ​​automatically analyze the content of the conversation and intervene at the appropriate timing. Methods of intervening at the appropriate timing include, but are not limited to, timing adjustment based on conversation content analysis and intervention based on specific cues, for example. For example, the intervention unit can have the AI ​​automatically analyze the content of the conversation using timing adjustment based on conversation content analysis and intervene at the appropriate timing. Alternatively, the intervention unit can have the AI ​​automatically analyze the content of the conversation using intervention based on specific cues and intervene at the appropriate timing. Alternatively, the intervention unit can combine timing adjustment based on conversation content analysis and intervention based on specific cues to intervene at the appropriate timing. For example, the intervention unit can automatically analyze the content of the conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. Alternatively, the AI ​​can automatically analyze the content of the conversation using intervention based on specific cues and intervene at the appropriate timing. Alternatively, the AI ​​can combine timing adjustment based on conversation content analysis and intervention based on specific cues to automatically analyze the content of the conversation and intervene at the optimal timing. This allows the AI ​​to automatically analyze the content of the conversation and intervene at the appropriate time, thereby reducing the risk of fraud. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without using AI. For example, the intervention unit may use an AI model that inputs conversation content data and intervenes at an appropriate timing to adjust the timing of intervention.

[0101] At the time of intervention, the intervention unit can select an optimal intervention method by referring to past intervention history. Methods of referring to past intervention history include, but are not limited to, storing and searching intervention logs and analyzing historical data. For example, the intervention unit analyzes past intervention history by storing and searching intervention logs and selecting the most effective intervention method. The intervention unit can also prioritize the use of intervention methods to which the user has responded in the past by analyzing historical data. The intervention unit can also learn the user's response pattern by combining storing and searching intervention logs and analyzing historical data and propose the optimal intervention method. For example, the intervention unit analyzes past intervention history by storing and searching intervention logs and selecting the most effective intervention method. The intervention unit prioritizes the use of intervention methods to which the user has responded in the past by analyzing historical data. The intervention unit learns the user's response pattern by combining storing and searching intervention logs and analyzing historical data and proposes the optimal intervention method. In this way, the optimal intervention method can be selected by referring to past intervention history. Some or all of the above-described processing in the intervention unit may be performed, for example, using AI or without AI. For example, the intervention unit can select an intervention method using an AI model that inputs past intervention history data and selects the optimal intervention method.

[0102] The intervention unit may add a function that allows family members or related parties to remotely intervene during intervention. Examples of functions that allow remote intervention include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The intervention unit may also use a voice call to allow related parties to remotely intervene and support the elderly. The intervention unit may also configure the system to allow family members or related parties to remotely intervene using chat. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The related parties may use a voice call to remotely intervene and support the elderly. The system may be configured to allow family members or related parties to remotely intervene using chat. This allows family members or related parties to remotely intervene, enabling a rapid response. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may allow family members or related parties to remotely intervene using an AI model that sets up a video call, voice call, or chat for remote intervention.

[0103] The intervention unit can estimate the user's emotion and adjust the timing of intervention based on the estimated user's emotion. Methods for adjusting the timing of intervention include, but are not limited to, timing adjustment based on an emotion score and adjustment based on a specific event. For example, the intervention unit estimates the user's emotion using timing adjustment based on an emotion score and adjusts the timing of intervention. The intervention unit can also estimate the user's emotion using adjustment based on a specific event and adjust the timing of intervention. The intervention unit can also adjust the timing of intervention by combining timing adjustment based on an emotion score and adjustment based on a specific event. For example, if the user is nervous, the intervention unit advances the timing of intervention to respond quickly. If the user is relaxed, the intervention unit returns the timing of intervention to normal and provides natural intervention. If the user is excited, the intervention timing can be adjusted to avoid an overreaction by combining timing adjustment based on an emotion score and adjustment based on a specific event. This allows for more appropriate intervention by adjusting the timing of intervention based on the user's emotion. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or without AI. For example, the intervention unit can adjust the timing of intervention using an AI model that takes the user's emotional data as input and adjusts the timing of intervention.

[0104] When intervening, the intervention unit can automatically analyze the tone and speed of the conversation using AI and intervene at the appropriate time. Methods for analyzing the tone and speed of the conversation include, but are not limited to, extraction of speech features and tone analysis algorithms. For example, the intervention unit can analyze the tone of the conversation in real time using extraction of speech features and intervene when there is a high possibility of fraud. The intervention unit can also analyze the tone of the conversation in real time using a tone analysis algorithm and intervene at the appropriate time. The intervention unit can also analyze the speed of the conversation in real time and intervene at the appropriate time. For example, the intervention unit can analyze the tone of the conversation in real time using extraction of speech features and intervene when there is a high possibility of fraud. The intervention unit can analyze the tone of the conversation in real time using a tone analysis algorithm and intervene at the appropriate time. The intervention unit can analyze the speed of the conversation in real time and intervene at the appropriate time. In this way, the AI ​​can automatically analyze the tone and speed of the conversation and intervene at the appropriate time, thereby reducing the risk of fraud. Some or all of the above-mentioned processing in the intervention unit may be performed using AI, for example, or without AI. For example, the intervention unit can adjust the timing of intervention using an AI model that takes data on the tone and speed of conversation as input and intervenes at the appropriate time.

[0105] The intervention unit may add a function that allows family members and related parties to intervene remotely. Examples of functions that allow remote intervention include, but are not limited to, video calls, voice calls, and chats. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The intervention unit may also use a voice call to allow related parties to remotely intervene and support the elderly. The intervention unit may also configure the system to allow family members and related parties to remotely intervene using chat. For example, the intervention unit may use a video call to allow family members to remotely participate in a conversation and warn of possible fraud. The related parties may use a voice call to remotely intervene and support the elderly. The system may be configured to allow family members and related parties to remotely intervene using chat. This allows family members and related parties to intervene remotely, enabling a rapid response. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without AI. For example, the intervention unit may allow family members and related parties to remotely intervene using an AI model that sets up a video call, voice call, or chat for remote intervention.

[0106] The intervention unit allows the AI ​​to automatically analyze the content of a conversation and intervene at an appropriate time. Methods of intervening at an appropriate time include, but are not limited to, timing adjustment based on conversation content analysis and intervening based on a specific cue. For example, the intervention unit allows the AI ​​to automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene at an appropriate time. The intervention unit can also allow the AI ​​to automatically analyze the content of a conversation using intervention based on a specific cue and intervene at an appropriate time. The intervention unit can also intervene at an appropriate time by combining timing adjustment based on conversation content analysis and intervention based on a specific cue. For example, the intervention unit allows the AI ​​to automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. The AI ​​can automatically analyze the content of a conversation using intervention based on a specific cue and intervene at an appropriate time. The AI ​​can automatically analyze the content of a conversation and intervene at an optimal time by combining timing adjustment based on conversation content analysis and intervention based on a specific cue. This allows the AI ​​to automatically analyze the content of a conversation and intervene at an appropriate time, thereby reducing the risk of fraud. Some or all of the above-described processing in the intervention unit may be performed using, for example, AI, or may be performed without using AI. For example, the intervention unit may use an AI model that inputs conversation content data and intervenes at an appropriate timing to adjust the timing of intervention. === Hard Collateral 1-1 === For example, the voice recognition unit is realized by the microphone 38B and the control unit 46A of the smart device 14. For example, the analysis unit is realized by the specific processing unit 290 of the data processing device 12. For example, the alert unit is realized by the control unit 46A of the smart device 14 or the specific processing unit 290 of the data processing device 12. For example, the intervention unit is realized by the control unit 46A of the smart device 14 or the specific processing unit 290 of the data processing device 12. === Hard Collateral 1-2 === For example, the voice recognition unit is realized by the microphone 238 and the control unit 46A of the smart glasses 214. For example, the analysis unit is realized by the specific processing unit 290 of the data processing device 12. For example, the alert unit is realized by the control unit 46A of the smart glasses 214 or the specific processing unit 290 of the data processing device 12. For example, the intervention unit is realized by the control unit 46A of the smart glasses 214 or the specific processing unit 290 of the data processing device 12. === Hard Collateral 1-3 === For example, the voice recognition unit is realized by the microphone 238 and the control unit 46A of the headset type terminal 314. For example, the analysis unit is realized by the specific processing unit 290 of the data processing device 12. For example, the alert unit is realized by the control unit 46A of the headset type terminal 314 or the specific processing unit 290 of the data processing device 12. For example, the intervention unit is realized by the control unit 46A of the headset type terminal 314 or the specific processing unit 290 of the data processing device 12. === Hard Collateral 1-4 === For example, the voice recognition unit is realized by the microphone 238 and the control unit 46A of the robot 414. For example, the analysis unit is realized by the specific processing unit 290 of the data processing device 12. For example, the alert unit is realized by the control unit 46A of the robot 414 or the specific processing unit 290 of the data processing device 12. For example, the intervention unit is realized by the control unit 46A of the robot 414 or the specific processing unit 290 of the data processing device 12.

[0107] The system according to the embodiment is not limited to the above-described example, and various modifications are possible, for example, as follows.

[0108] The analysis unit can estimate the user's emotions and adjust the reliability of the analysis results based on the estimated user emotions. For example, if the user is nervous, the reliability of the analysis results can be increased to provide more accurate results. If the user is relaxed, the reliability of the analysis results can be returned to normal to perform a natural analysis. Furthermore, if the user is excited, the reliability of the analysis results can be adjusted to reduce erroneous judgments. In this way, by adjusting the reliability of the analysis results based on the user's emotions, more accurate analysis results can be provided.

[0109] During speech recognition, the speech recognition unit can learn the characteristics of a speaker's voice and use an individually optimized speech recognition model. For example, it can use acoustic feature extraction to learn the tone and pitch of a speaker's voice and generate an individually optimized speech recognition model. It can also use machine learning model training to learn the pronunciation habits and speed of a speaker and use an individually optimized speech recognition model. Furthermore, it can continuously learn the characteristics of a speaker's voice and update the speech recognition model. As a result, by learning the characteristics of a speaker's voice and using an individually optimized speech recognition model, the accuracy of speech recognition is improved.

[0110] The alert unit can estimate the user's emotions and adjust the intensity of the alert based on the estimated user emotions. For example, if the user is nervous, the alert intensity can be increased to quickly attract attention. If the user is relaxed, the alert intensity can be returned to normal to provide a natural notification. Furthermore, if the user is excited, the alert intensity can be adjusted to avoid an overreaction. This allows for more appropriate alerts by adjusting the alert intensity based on the user's emotions.

[0111] The intervention unit allows the AI ​​to automatically analyze the content of a conversation and intervene at the appropriate time. For example, the AI ​​can automatically analyze the content of a conversation using timing adjustment based on conversation content analysis and intervene when there is a high possibility of fraud. The AI ​​can also automatically analyze the content of a conversation and intervene at the appropriate time using intervention based on specific cues. Furthermore, timing adjustment based on conversation content analysis and intervention based on specific cues can be combined to intervene at the optimal time. This allows the AI ​​to automatically analyze the content of a conversation and intervene at the appropriate time, thereby reducing the risk of fraud.

[0112] The speech recognition unit can add a filtering function that automatically removes background noise during speech recognition. For example, a noise canceling algorithm can be used to automatically remove background noise during speech recognition. It can also use filtering technology to remove noise in specific frequency bands during speech recognition. Furthermore, it can learn environmental sounds and automatically remove specific patterns of noise. This automatically removes background noise, improving the accuracy of speech recognition.

[0113] During the analysis, the analysis unit can analyze the tone and speed of the conversation to determine the possibility of fraud. For example, speech feature extraction can be used to analyze the tone of the conversation to determine the possibility of fraud. A tone analysis algorithm can also be used to analyze the tone of the conversation in real time to determine the possibility of fraud. Furthermore, the speed of the conversation can also be analyzed in real time to determine the possibility of fraud. Thus, by analyzing the tone and speed of the conversation, the possibility of fraud can be more accurately determined.

[0114] The alert unit can combine multiple notification methods to issue an alert when an alert occurs. For example, a smartphone notification and email can be sent simultaneously, allowing the user to be notified quickly. It can also combine a voice alert with a smartphone notification to notify the user both visually and audibly. It can also combine an email and a voice alert to reliably notify the user of the alert. In this way, by combining multiple notification methods, it is possible to quickly and reliably notify the user of the alert.

[0115] The intervention unit can estimate the user's emotions and adjust the intervention method based on the estimated user's emotions. For example, if the user is nervous, the intervention unit can use a calm tone. If the user is relaxed, the intervention unit can use a normal tone. If the user is excited, the intervention unit can use a calm tone. This allows for more appropriate intervention by adjusting the intervention method based on the user's emotions.

[0116] During analysis, the analysis unit can improve analysis accuracy by learning not only past fraud cases but also the latest fraud techniques. For example, it can improve analysis accuracy by collecting and learning from the latest fraud datasets to learn not only past fraud cases but also the latest fraud techniques. It can also improve analysis accuracy by using an online learning algorithm to learn the latest fraud techniques in real time. Furthermore, it can also update the analysis model by combining past fraud cases and the latest fraud techniques. This improves analysis accuracy by learning the latest fraud techniques.

[0117] The intervention unit can estimate the user's emotions and adjust the timing of intervention based on the estimated user emotions. For example, if the user is nervous, the timing of intervention can be accelerated to respond quickly. Also, if the user is relaxed, the timing of intervention can be returned to normal to provide natural intervention. Furthermore, if the user is excited, the timing of intervention can be adjusted to avoid an excessive reaction. Thus, by adjusting the timing of intervention based on the user's emotions, more appropriate intervention is possible.

[0118] The processing flow of the second embodiment will be briefly explained below.

[0119] Step 1: The speech recognition unit recognizes the voice conversation in real time, for example, by using a deep learning model to convert speech to text, and converts what the elderly person is saying on the phone into text in real time. Step 2: The analysis unit analyzes the speech recognized by the speech recognition unit. For example, it uses a model trained on past fraud cases to determine the possibility of fraud, and if the speech contains phrases such as "Please transfer money" or "Please respond quickly," it determines that there is a high possibility of fraud. Step 3: The alert unit issues an alert if a fraud risk is detected based on the results of the analysis by the analysis unit. For example, it can send a notification to the elderly person's smartphone, send an email, or send a voice warning, or send an email to their family. Step 4: The intervention unit intervenes in the conversation when an alert is issued by the alert unit. For example, the AI ​​can automatically interrupt the conversation and warn, "This conversation may be fraudulent." Family members can also remotely join the conversation and provide support to the elderly.

[0120] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0121] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> Examples of generative AIs include the data generation model 58, such as a neural network model (e.g., a neural network model), and a neural network model (e.g., a neural network model). The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating speech, text data indicating text, and image data indicating an image is also input to the data generation model 58. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specification processing unit 290 performs the above-mentioned specification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model so as to output an inference result from a prompt that does not include an instruction. In this case, the data generation model 58 can output an inference result from a prompt that does not include an instruction. The data processing device 12 and the like include multiple types of data generation models 58, and the data generation model 58 includes AIs other than the generative AI. The AI ​​other than the generative AI may be, for example, linear regression, logistic regression, decision tree, random forest, support vector machine (SVM), k-means clustering, convolutional neural network (CNN), recurrent neural network (RNN), generative adversarial network (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. The AI ​​may also be an AI agent. When the processes of each of the above-mentioned parts are performed by AI, the processes may be performed in part or entirely by AI, but are not limited to these examples. The processes performed by AI, including the generative AI, may be replaced with rule-based processes.

[0122] Furthermore, the processing by the data processing system 10 described above is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart device 14 or an external device, and the smart device 14 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0123] The correspondence between each part and the device or control part is not limited to the example described above, and various modifications are possible.

[0124] [Second embodiment] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0125] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0126] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN and / or a LAN.

[0127] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0128] The microphone 238 receives instructions and the like from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to instructions from the processor 46.

[0129] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the user's surroundings (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0130] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0131] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0132] 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.

[0133] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate a user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotion, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, the estimation and prediction of emotion also includes, for example, emotion analysis.

[0134] In the smart glasses 214, the specific processing is performed by the processor 46. A specific processing program 60 is stored in the storage 50. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as the control unit 46A in accordance with the specific processing program 60 executed on the RAM 48. The smart glasses 214 also have a data generation model and an emotion identification model similar to the data generation model 58 and the emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0135] Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain a processing result (such as a prediction result) using the data generation model 58. Furthermore, the data processing device 12 may be a server device, or may be a terminal device (for example, a mobile phone, a robot, a home appliance, etc.) owned by a user.

[0136] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0137] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 receives a prompt containing an instruction, as well as inference data such as voice data representing speech, text data representing text, and image data representing an image. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model so as to output an inference result from a prompt that does not include an instruction. In this case, the data generation model 58 can output an inference result from a prompt that does not include an instruction. The data processing device 12 and the like include multiple types of data generation models 58, and the data generation model 58 includes AIs other than the generative AI. The AI ​​other than the generative AI may be, for example, linear regression, logistic regression, decision tree, random forest, support vector machine (SVM), k-means clustering, convolutional neural network (CNN), recurrent neural network (RNN), generative adversarial network (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. The AI ​​may also be an AI agent. When the processes of each of the above-mentioned parts are performed by AI, the processes may be performed in part or entirely by AI, but are not limited to these examples. The processes performed by AI, including the generative AI, may be replaced with rule-based processes.

[0138] The data processing system 210 according to the second embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 210 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or an external device, etc., and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or an external device, etc.

[0139] The correspondence between each part and the device or control part is not limited to the example described above, and various modifications are possible.

[0140] [Third embodiment] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0141] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0142] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN and / or a LAN.

[0143] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0144] The microphone 238 receives instructions and the like from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to instructions from the processor 46.

[0145] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the user's surroundings (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0146] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0147] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0148] 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.

[0149] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate a user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotion, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, the estimation and prediction of emotion also includes, for example, emotion analysis.

[0150] In the headset type terminal 314, the identification process is performed by the processor 46. A identification program 60 is stored in the storage 50. The processor 46 reads the identification program 60 from the storage 50 and executes the read identification program 60 on the RAM 48. The identification process is realized by the processor 46 operating as a control unit 46A in accordance with the identification program 60 executed on the RAM 48. Note that the headset type terminal 314 has a data generation model and an emotion identification model similar to the data generation model 58 and the emotion identification model 59, and can also perform processing similar to that of the identification processing unit 290 using these models.

[0151] Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain a processing result (such as a prediction result) using the data generation model 58. Furthermore, the data processing device 12 may be a server device, or may be a terminal device (for example, a mobile phone, a robot, a home appliance, etc.) owned by a user.

[0152] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0153] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 receives a prompt containing an instruction, as well as inference data such as voice data representing speech, text data representing text, and image data representing an image. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model so as to output an inference result from a prompt that does not include an instruction. In this case, the data generation model 58 can output an inference result from a prompt that does not include an instruction. The data processing device 12 and the like include multiple types of data generation models 58, and the data generation model 58 includes AI other than the generative AI. The AI ​​other than the generative AI may be, for example, linear regression, logistic regression, decision tree, random forest, support vector machine (SVM), k-means clustering, convolutional neural network (CNN), recurrent neural network (RNN), generative adversarial network (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. The AI ​​may also be an AI agent. When the processes of each of the above-mentioned parts are performed by AI, the processes may be performed in part or entirely by AI, but are not limited to these examples. The processes performed by AI, including the generative AI, may be replaced with rule-based processes.

[0154] The data processing system 310 according to the third embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 310 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the headset type terminal 314, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the headset type terminal 314. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the headset type terminal 314 or an external device, etc., and the headset type terminal 314 acquires or collects information required for processing from the data processing device 12 or an external device, etc.

[0155] The correspondence between each part and the device or control part is not limited to the example described above, and various modifications are possible.

[0156] [Fourth embodiment] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0157] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0158] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN and / or a LAN.

[0159] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[0160] The microphone 238 receives instructions and the like from the user by receiving voice uttered by the user. The microphone 238 captures the voice uttered by the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to instructions from the processor 46.

[0161] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS image sensor or a CCD image sensor, and captures images of the user's surroundings (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0162] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0163] The control 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 emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[0164] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0165] 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.

[0166] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate a user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotion, including estimation and prediction of the user's emotion, but is not limited to these examples. Furthermore, the estimation and prediction of emotion also includes, for example, emotion analysis.

[0167] In the robot 414, the processor 46 performs the identification process. The storage 50 stores the identification program 60. The processor 46 reads the identification program 60 from the storage 50 and executes the read identification program 60 on the RAM 48. The identification process is realized by the processor 46 operating as the control unit 46A in accordance with the identification program 60 executed on the RAM 48. The robot 414 also has a data generation model and an emotion identification model similar to the data generation model 58 and the emotion identification model 59, and can perform the same process as the identification processing unit 290 using these models.

[0168] Note that a device other than the data processing device 12 may have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 communicates with the server device having the data generation model 58 to obtain a processing result (such as a prediction result) using the data generation model 58. Furthermore, the data processing device 12 may be a server device, or may be a terminal device (for example, a mobile phone, a robot, a home appliance, etc.) owned by a user.

[0169] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0170] The data generation model 58 is a so-called generative AI. An example of the data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 receives a prompt containing an instruction, as well as inference data such as voice data representing speech, text data representing text, and image data representing an image. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The identification processing unit 290 performs the above-mentioned identification processing using the data generation model 58. The data generation model 58 may be a fine-tuned model so as to output an inference result from a prompt that does not include an instruction. In this case, the data generation model 58 can output an inference result from a prompt that does not include an instruction. The data processing device 12 and the like include multiple types of data generation models 58, and the data generation model 58 includes AI other than the generative AI. The AI ​​other than the generative AI may be, for example, linear regression, logistic regression, decision tree, random forest, support vector machine (SVM), k-means clustering, convolutional neural network (CNN), recurrent neural network (RNN), generative adversarial network (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. The AI ​​may also be an AI agent. When the processes of each of the above-mentioned parts are performed by AI, the processes may be performed in part or entirely by AI, but are not limited to these examples. The processes performed by AI, including the generative AI, may be replaced with rule-based processes.

[0171] The data processing system 410 according to the fourth embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 410 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but may also be executed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Furthermore, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or an external device, etc., and the robot 414 acquires or collects information required for processing from the data processing device 12 or an external device, etc.

[0172] The correspondence between each part and the device or control part is not limited to the example described above, and various modifications are possible.

[0173] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0174] FIG. 9 illustrates an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and behaviors arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion encompasses both emotions and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[0175] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[0176] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[0177] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is expressed, and when they approach the ideal, a state of pleasure is expressed. Emotions can also be created for robots, cars, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is expressed, and when they approach the ideal, a state of pleasure is expressed. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on speech emotion recognition and brain physiological signal analysis systems for emotions, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[0178] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[0179] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[0180] In the above embodiment, an example was given in which a specific process is performed by one computer 22, but the technology disclosed herein is not limited to this, and distributed processing of the specific process may be performed by multiple computers including computer 22.

[0181] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[0182] 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.

[0183] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[0184] The hardware resource for executing a specific process can be any of the following types of processors: A processor, for example, is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. A processor also includes a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[0185] The hardware resource that executes the specific process may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific process may be a single processor.

[0186] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[0187] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[0188] In the above example, the first to fourth embodiments have been described separately, but some or all of these embodiments may be combined. The smart device 14, smart glasses 214, headset terminal 314, and robot 414 are merely examples, and they may be combined, or other devices may be used. In the above example, the first and second embodiments have been described separately, but they may be combined.

[0189] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[0190] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[0191] [Explanation of symbols]

[0192] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot

Claims

1. a voice recognition unit that recognizes voice conversations in real time; an analysis unit that analyzes the voice recognized by the voice recognition unit; an alert unit that issues an alert when a risk of fraud is detected based on the analysis result by the analysis unit; an intervention unit that intervenes in the conversation when an alert is issued by the alert unit; A system characterized by:

2. The voice recognition unit Convert speech to text using deep learning models 2. The system of claim 1.

3. The analysis unit Determine the likelihood of fraud using a model trained on past fraud cases 2. The system of claim 1.

4. The alert unit Notifications on your smartphone, email, or voice alerts 2. The system of claim 1.

5. The intervention unit is Family members can join in the conversation remotely and support the elderly 2. The system of claim 1.

6. The voice recognition unit Estimate the user's emotions and adjust the accuracy of speech recognition based on the estimated user emotions.

2. The system of claim 1.

7. The voice recognition unit Add a filtering function to automatically remove background noise during voice recognition.

2. The system of claim 1.

Images