system

Abstract

Provide a system. 【Solution means】 Means for collecting a user's voice and converting it into digital data, Means for extracting features from the digital data, Means for performing sentiment analysis on the feature-extracted data, Means for generating an optimal response based on the analysis result, Means for synthesizing the generated response as voice, Means for playing back the synthesized voice to the user, Means for generating a response that empathizes with the emotions of the elderly based on the result of the sentiment analysis, A system including the above.

Description

Technical Field

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

Background Art

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

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In conventional voice interface systems, it has been difficult to accurately analyze the user's emotions and provide a response according to the emotions, resulting in the problem that the user experience is not improved. In particular, in order to realize a natural response that empathizes with emotions, it is necessary to integrate advanced voice analysis technology and natural language processing technology. However, it has been difficult to implement these technologies efficiently, and a simple and effective method has been demanded.

Means for Solving the Problems

[0005] This invention provides a means for collecting user voice in real time and converting it into digital data. A key feature is the extraction of emotional characteristics from the voice data using Mel-frequency cepstrum coefficients, followed by emotion analysis using a machine learning model. Furthermore, the system generates an optimal response based on the analysis results and uses speech synthesis technology to reproduce an emotionally rich response to the user. This enables the provision of natural responses that resonate with the user's emotions, resulting in a superior user experience.

[0006] A "user" refers to a person who uses the system to input voice commands.

[0007] "Voice collection" refers to the process of capturing the voice spoken by the user using a microphone or other device.

[0008] "Digital data" refers to data obtained by converting collected analog audio signals into a format that can be processed by a computer.

[0009] "Feature extraction" refers to the process of extracting information necessary to analyze specific patterns or characteristics from audio data.

[0010] "Emotion analysis" refers to the process of analyzing feature-extracted data to identify the emotions contained within it.

[0011] "Response generation" refers to the process of constructing the optimal response for the user based on the results of sentiment analysis.

[0012] "Speech synthesis" refers to the process of reconstructing generated responses as speech and making them audible to the user.

[0013] "Playback" refers to the process of transmitting synthesized audio audibly to the user through speakers or other means.

[0014] A "system" refers to a series of processes or devices that integrate the above-mentioned means and function as a whole. [Brief explanation of the drawing]

[0015] [Figure 1] It is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] It is a conceptual diagram showing an example of the main functions of a data processing device and a smart device according to the first embodiment. [Figure 3] It is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] It is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] It is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] It is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] It is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] It is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] Shows an emotion map to which multiple emotions are mapped. [Figure 10] Shows an emotion map to which multiple emotions are mapped. [Figure 11] It is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Example 2 when the emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when the emotion engine is combined.

Embodiments for Carrying Out the Invention

[0016] An example of an embodiment of the system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0017] First, the terms used in the following description will be explained.

[0018] In the following embodiments, the numbered processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Also, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), and the like.

[0019] In the following embodiments, the numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0020] In the following embodiments, the numbered storage is one or more non-volatile storage devices that store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes, and the like.

[0021] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0023] [First Embodiment]

[0024] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0025] As shown in Figure 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0026] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0028] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0031] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0032] As shown in Figure 2, in the data processing device 12, a specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0033] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0034] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0035] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0036] The empathetic AI agent system according to the present invention collects voice from the user in real time and emotionally analyzes the voice to provide the user with a more appropriate and emotionally rich response. An embodiment of this system is shown below.

[0037] First, the terminal collects the user's voice through a voice input device. This voice is converted into digital data in real time. Next, the terminal extracts features from this digital data and sends them to the server as samples. In particular, feature extraction is performed from the voice data using Mel-frequency cepstrum coefficients (MFCCs).

[0038] The server uses the received audio feature data to run machine learning algorithms and analyze the user's emotions. A pre-trained deep learning model is used for this, and emotion labels are assigned to the analysis results. Based on these emotion labels, the server utilizes natural language processing techniques to generate an appropriate response in text format.

[0039] The generated response text is again converted into speech data by the server using speech synthesis technology. This speech data is synthesized with attention to emotional nuances and is ultimately sent to the terminal.

[0040] The device plays the received synthesized speech data to the user through its speaker. This allows the user to receive responses that are natural and relatable, tailored to their own emotions.

[0041] As a concrete example, consider a scenario where a user says, "I'm tired today." In this case, the voice is collected by the device, and emotion analysis identifies the emotion as "fatigue." The response generation process creates a message such as, "You've worked hard. Please get plenty of rest," and the voice is synthesized and played back to the user. As a result, the user receives a response that reflects their own emotions, making the interaction with the system more natural. Thus, this invention realizes dialogue with an AI agent that matches the user's emotions, thereby improving the quality of life.

[0042] The following describes the processing flow.

[0043] Step 1:

[0044] The device collects the user's voice through the microphone. The collected voice is converted into a digital signal, and noise is filtered out to produce clean audio data.

[0045] Step 2:

[0046] The terminal divides the digital signal into short segments and extracts speech features from each segment. Here, Mel-frequency cepstrum coefficients (MFCCs) are used to calculate the speech features.

[0047] Step 3:

[0048] The terminal sends the extracted feature data to the server. This transmission may involve data compression to optimize network efficiency.

[0049] Step 4:

[0050] The server uses the received audio feature data to perform sentiment analysis using a machine learning model. The analysis generates labels indicating emotions and identifies the user's emotional state.

[0051] Step 5:

[0052] The server generates response text that matches the original user statement based on the sentiment analysis results. Natural language processing techniques are used to select wording that reflects the user's emotions.

[0053] Step 6:

[0054] The server converts the generated response text into speech data using a speech synthesis engine. During this process, it adjusts intonation, pitch, and speed to express emotional nuances.

[0055] Step 7:

[0056] The server sends the synthesized audio data to the terminal. This transmission also involves data compression as needed.

[0057] Step 8:

[0058] The device plays the received audio data through its speaker. In this way, it provides the user with an emotionally resonant voice response.

[0059] (Example 1)

[0060] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0061] Traditional interaction technologies face the challenge of generating empathetic responses to user emotions in real time. Furthermore, if the generated responses do not adequately reflect the nuances of those emotions, communication with the user tends to become strained.

[0062] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0063] In this invention, the server includes means for acquiring the user's voice and converting it into an electrical signal, means for reading features from the electrical signal, and means for evaluating emotions using the read features. This enables the server to appropriately evaluate the user's emotions in real time, quickly generate empathetic responses based on those evaluations, and facilitate natural communication with the user.

[0064] "Voice" refers to the audio signals emitted by the user, which are collected as acoustic information.

[0065] An "electrical signal" is a format used for information processing, which is the conversion of sound into digital data.

[0066] A "feature" is a specific element of data extracted from an audio signal that contains information necessary for sentiment evaluation.

[0067] "Emotion" refers to the psychological state a user is in, and is identified through their vocal characteristics.

[0068] "Evaluation" is the process of determining the user's emotions based on features extracted from their voice.

[0069] "Response" refers to an audio or text message generated based on the user's sentiment evaluation.

[0070] "Synthesis" refers to the process of generating speech from text or data, and is a technology for providing voice responses to users.

[0071] "Output" refers to playing the generated audio through a device such as a speaker.

[0072] The emotion-empathizing AI agent system according to this invention uses multiple hardware and software components to provide natural interaction based on the user's emotions.

[0073] First, the terminal uses a smartphone or computer equipped with a microphone as an audio input device. The user's voice is picked up by the microphone as an analog signal and converted into an electrical signal on the spot. The electrical signal is sampled over a specific time interval through signal processing technology, and features are extracted as digital data. For this process, Mel-frequency cepstrum coefficients (MFCCs) are used as a method for frequency conversion.

[0074] Next, the server receives feature data sent from the terminal and uses a pre-trained learning algorithm to evaluate the user's emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used for emotion evaluation. As a result of the evaluation, emotion labels such as "joy," "sadness," and "anger" are assigned.

[0075] Based on the received emotion label, the server uses natural language processing techniques to generate an empathetic response. This response generation utilizes a generative AI model, which generates appropriate text in response to prompt text. For example, the prompt text "What should I say if the user's emotion is 'fatigue'?" is input to the model.

[0076] The generated text responses are converted into audio data using speech synthesis technology. Technologies such as WaveNet are used for speech synthesis to achieve natural and fluent expression. The audio data is then sent back to the terminal.

[0077] Ultimately, the device outputs synthesized speech data to the user through its speaker. This allows the user to naturally receive empathetic responses from the system, resulting in a more human-like interface.

[0078] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0079] Step 1:

[0080] The user emits sound through the microphone. This sound is input to the device and captured as an analog signal. Next, the device converts the analog signal into an electrical signal. The resulting electrical signal becomes the input for the audio data.

[0081] Step 2:

[0082] The terminal converts this electrical signal into digital audio data and samples it over a specific time interval. From the sampled digital data, it performs frequency conversion to extract Mel-frequency cepstrum coefficients (MFCCs). The MFCCs become the output feature data, representing important characteristics of the audio as numerical data.

[0083] Step 3:

[0084] The terminal sends the extracted MFCC data to the server as sample data. The server receives this data and uses it as input data for sentiment analysis. This transmitted feature data is the input for the next process.

[0085] Step 4:

[0086] The server uses the received feature data to run a deep learning model and evaluate emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used here. This outputs emotion labels. For example, emotions such as "joy," "anger," and "sadness" are analyzed.

[0087] Step 5:

[0088] The server generates empathetic response text using natural language processing techniques based on emotion labels. This process utilizes a generative AI model, which is input with prompts to output appropriate text. For example, a prompt might ask, "What should I say if the user's emotion is 'fatigue'?"

[0089] Step 6:

[0090] The server uses speech synthesis technology to convert the generated response text into speech. The generated speech data is synthesized using models such as WaveNet to produce output that reflects emotions.

[0091] Step 7:

[0092] The device receives synthesized speech data sent from the server and plays it through the speaker. Through this outputted speech, the user can receive emotionally resonant responses.

[0093] (Application Example 1)

[0094] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0095] Communicating with the elderly requires understanding their emotions appropriately and providing empathetic responses. However, conventional techniques are insufficient in adequately addressing the unique emotional states of the elderly. Therefore, there is a need for more effective communication methods to improve the quality of life for the elderly.

[0096] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0097] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, and means for generating responses that are empathetic to the emotions of elderly people based on the results of emotion analysis. This makes it possible to understand the unique emotional state of elderly people and to communicate appropriately with them.

[0098] "Sound" refers to the audio signal received from the user, which is the input data that the system uses to convert it into digital data and analyze it.

[0099] "Digital data" refers to data obtained by converting audio into a format that can be processed on a computer, and it serves as fundamental information used for feature extraction and sentiment analysis.

[0100] "Feature extraction" is the process of extracting specific information from audio data and deriving metrics necessary for data analysis. It is a crucial process that affects the accuracy of the analysis.

[0101] "Emotional analysis" is a process that analyzes the emotions contained in a user's utterances based on feature-extracted data and determines a specific emotional state.

[0102] "Response generation" is the process of creating appropriate text responses based on the results of sentiment analysis, enabling communication that is sensitive to the user's emotions.

[0103] "Speech synthesis" is the process of converting text-generated responses back into speech format, providing users with natural-sounding responses.

[0104] "Empathizing with the emotions of the elderly" means understanding the unique emotional states that elderly people exhibit, responding appropriately and reassuringly, and aiming to achieve comfortable communication.

[0105] To implement this invention, it is necessary to use a terminal that functions as a user interface and a server that performs voice data analysis. The terminal is equipped with a voice input device that collects the user's voice in real time and converts it into digital data.

[0106] The digital data converted by the terminal is sent to a server for feature extraction. The server uses software such as Python or LibROSA to extract Mel-frequency cepstrum coefficients. This allows the necessary features for sentiment analysis to be obtained from the audio data.

[0107] Sentiment analysis is performed on a server using a pre-trained deep learning model. This process utilizes machine learning frameworks such as TENSORFLOW® to assign labels indicating specific emotions to audio data. Subsequently, Hugging Face's Transformers are used to generate natural responses based on the analysis results.

[0108] The generated responses are converted into audio data using speech synthesis software such as Google® Text-to-Speech. The converted audio data is sent to the terminal and played back to the user. Through this series of processes, the system enables communication that is sensitive to the emotions of elderly people.

[0109] For example, if a user says, "My family isn't coming," the server can generate an empathetic response such as, "They are important family members, aren't they? Shall I think about what you'd like to talk about next time they visit?" An example of a prompt might be, "Generate a response that reflects the following emotion: Sadness - Appropriate comfort when the person you're talking to is sad."

[0110] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0111] Step 1:

[0112] The terminal collects the user's voice using an audio input device. The input is an analog audio signal, which is then converted into digital data. The converted digital data is then ready for transmission to the server.

[0113] Step 2:

[0114] The server processes the digital audio data received from the terminal and performs feature extraction. Specifically, it uses the LibROSA library to calculate Mel-frequency cepstrum coefficients. The input is digital audio data, and the output is feature data. This allows the characteristics of the audio to be quantified and used for sentiment analysis.

[0115] Step 3:

[0116] The server performs sentiment analysis based on feature data. A deep learning model (using TensorFlow) generates sentiment labels. The input is feature data, and the output is sentiment labels. These labels indicate the user's emotional state.

[0117] Step 4:

[0118] The server generates a response using natural language processing techniques based on the generated emotion labels. A text-based response is created using Hugging Face's Transformers. The input is the emotion label, and the output is the text response to be returned to the user.

[0119] Step 5:

[0120] The server converts text responses into speech data. It uses Google Text-to-Speech to synthesize speech from text. The input is the text response, and the output is synthesized speech data. This generates natural-sounding speech responses.

[0121] Step 6:

[0122] The terminal receives synthesized speech data from the server and plays it back to the user through the speaker. The input is synthesized speech data, and the output is the voice the user hears. This enables communication that is sensitive to the emotions of elderly people.

[0123] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0124] The emotion-empathizing AI agent system according to the present invention provides a new interactive system that combines an emotion engine to highly recognize the user's emotions.

[0125] First, the device uses its voice input function to collect the user's voice and converts the analog audio signal into digital data. In this process, the audio data is processed through noise filtering to make it clearer.

[0126] Next, the device extracts voice features from this digital data using Mel-frequency cepstrum coefficients (MFCCs), and in addition, collects user behavior data (e.g., gestures and behaviors), which it then sends to the emotion engine.

[0127] The server drives an emotion engine, integrating voice and behavioral data to analyze the user's emotions. A deep learning-based machine learning model is used for emotion recognition, which generates a variety of emotion labels. The emotion engine precisely identifies multiple emotional states, and the analysis results are stored on the server.

[0128] Based on the analysis results, the server utilizes natural language processing technology to generate responses that resonate with the user's emotions. The content and tone of the responses are dynamically adjusted based on the emotion recognition results. The generated response text is then converted into natural, emotionally expressive speech data by a speech synthesis engine.

[0129] Ultimately, the device plays synthesized speech to the user. This playback is done through the speaker, improving the user experience by providing responses appropriate to the situation the user is facing.

[0130] For example, if a user says "I'm sad today," the emotion engine immediately recognizes the emotion of "sadness." The server then generates a response of encouragement such as "You've had a tough day, please let me know if there's anything I can do to help," which is then played back via speech synthesis, providing the user with an empathetic interaction. In this way, the present invention provides a system that realizes a rich emotional experience for the user.

[0131] The following describes the processing flow.

[0132] Step 1:

[0133] The device collects the user's voice through a microphone. This voice is acquired as an analog signal and immediately converted into digital data. Furthermore, the voice data is processed using noise filtering to remove unwanted background noise and prepare it for voice analysis.

[0134] Step 2:

[0135] The device extracts features from digital audio data using Mel-frequency cepstrum coefficients (MFCCs). These features are numerical data that represent the characteristics of the voice and are used as basic data for speech recognition and sentiment analysis. The device also simultaneously collects user behavior data (e.g., gesture recognition via camera).

[0136] Step 3:

[0137] The terminal combines the extracted voice data and behavioral data and sends it to the server. The data is compressed during transmission to reduce network load.

[0138] Step 4:

[0139] The server analyzes the received data using an emotion engine. The emotion engine integrates voice and behavioral data using machine learning models to identify the user's emotional state. The identified emotions are categorized into specific labels, and the results are processed on the server.

[0140] Step 5:

[0141] The server applies a natural language processing algorithm based on the sentiment analysis results to generate the optimal response for the user. This response is designed with content and tone that resonates with the user's current emotions.

[0142] Step 6:

[0143] The server uses the generated response text to initiate the speech synthesis process. This process synthesizes speech that reflects emotional expression from the generated text, constructing the final audio file.

[0144] Step 7:

[0145] The server sends the synthesized audio to the terminal. The terminal also decodes the data as needed and prepares it to be delivered to the user in the optimal state.

[0146] Step 8:

[0147] The device plays the received synthesized speech through its speaker. This allows the user to hear responses in a voice and expression that reflects their own emotions, enabling them to enjoy a natural dialogue with the system.

[0148] (Example 2)

[0149] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0150] Conventional voice dialogue systems face the challenge of not being able to appropriately recognize user emotions and provide empathetic responses. Furthermore, it is necessary to perform more accurate emotion analysis by taking into account not only voice information but also behavioral information.

[0151] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0152] In this invention, the server includes means for acquiring user voice and converting it into digital information, means for extracting characteristics from the digital information, and means for acquiring user behavior and using it for emotion analysis. This enables integrated analysis of emotions from voice and behavior, and the generation of responses that are tailored to the user.

[0153] "User voice" refers to the human voice signal input into the system.

[0154] "Digital information" refers to data obtained by converting analog signals into a digital format.

[0155] "Characteristics" refer to distinctive elements or patterns contained in voice or behavioral data.

[0156] "Emotions" refer to information that indicates the user's psychological state.

[0157] A "response" is a message or action that a system returns to the user.

[0158] "Behavior" refers to information obtained from the user's gestures and actions.

[0159] The "frequency coefficient" is an index used to analyze the characteristics of audio data.

[0160] A "machine learning algorithm" is a technology that learns patterns from data and uses them to make predictions and classifications.

[0161] "Natural language processing technology" is a technology that enables computers to understand and generate human language.

[0162] In this invention, the user first inputs their voice into the system using a terminal equipped with an audio input device. The terminal is equipped with a high-performance microphone that can accurately capture the voice. The audio signal is converted from analog to digital, and noise is removed using digital signal processing technology. This results in clear audio data.

[0163] Next, the device uses signal processing software to extract speech characteristics by calculating Mel-frequency cepstrum coefficients (MFCCs) from the digitized speech data. It also captures user gestures and movements through cameras and motion sensors, collecting this as behavioral data. This consistent data, combined with speech, forms the basis for sentiment analysis.

[0164] The server uses a generative AI model powered by deep learning to evaluate integrated speech and behavioral data through sentiment analysis. This model is trained on a large amount of data and accurately identifies human emotions as multiple emotion labels. The identified emotions are stored on the server and updated as needed.

[0165] The server then uses natural language processing technology to generate a response that matches the user's emotional state. The generated response is adjusted in content and tone to be more natural and empathetic to the user's feelings. For example, if the user says, "I'm sad today," the system recognizes the emotion of "sadness" and generates a response such as, "You've had a tough day, please let me know if there's anything I can do to help."

[0166] The generated text is converted into expressive speech through a speech synthesis engine. The device then plays this synthesized speech through its speaker, delivering it directly to the user. This process allows the user to have an experience that feels like interacting with a human, rather than a robotic one.

[0167] Example of a prompt

[0168] "Today was very busy, and I'm a little tired. Could you tell me how to relax?"

[0169] "I'm feeling a little anxious about the new project. Could you give me some advice?"

[0170] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0171] Step 1:

[0172] The user inputs their voice into the device via the microphone. The device receives this voice as an analog signal and converts it to a digital signal using an analog-to-digital converter. This converted digital audio data is then subjected to a noise reduction filter to reduce background noise and obtain a clear result. This then becomes the input for the next step.

[0173] Step 2:

[0174] The device uses the cleared digital audio data to extract Mel-frequency cepstrum coefficients (MFCCs) using a signal processing algorithm. This analyzes the characteristics of the audio, which are then output as feature data. In addition, the camera and motion sensors capture the user's gestures and movements, which are stored as behavioral data. This data is then transmitted to the emotion analysis engine.

[0175] Step 3:

[0176] The server receives voice feature data and behavioral data sent from the terminal. Using a generative AI model, this data is integrated and sentiment analysis is performed. This analysis classifies the data into numerous sentiment labels, allowing for an understanding of the user's emotional state. This is crucial input for the next step.

[0177] Step 4:

[0178] The server uses natural language processing technology to generate user-specific responses based on the sentiment analysis results. The content and tone of the responses are adjusted according to the analyzed sentiment labels. The generated responses are output in text format.

[0179] Step 5:

[0180] The terminal receives generated text sent from the server and converts it into speech data using a speech synthesis engine. Speech synthesis produces speech with natural and emotional intonation, which becomes the final output.

[0181] Step 6:

[0182] The device plays synthesized speech to the user via its speaker. At this point, the user receives empathetic responses from the system in voice, completing the conversational experience.

[0183] (Application Example 2)

[0184] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0185] In elderly care settings, it is crucial to accurately understand changes in emotions and conditions, and to automatically provide empathetic dialogue and instructions accordingly. A lack of emotional understanding can lead to a decline in the satisfaction and quality of life of elderly individuals. A system is needed to address this situation while simultaneously reducing the burden on caregivers.

[0186] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0187] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, means for performing sentiment analysis on the extracted feature data and behavioral data, means for generating empathetic responses based on the analysis results, and means for synthesizing the generated responses as voice and making suggestions based on a specific context. This makes it possible to accurately grasp the emotional state of the person receiving care and provide empathetic and appropriate responses.

[0188] A "user" refers to a person who uses a system and interacts with it.

[0189] "Speech" refers to a waveform signal containing language emitted by the user.

[0190] "Digital data" is a data format that quantifies analog audio signals.

[0191] "Feature extraction" is the process of extracting meaningful patterns or information from digital data.

[0192] "Behavioral data" refers to information about a user's actions, gaze, gestures, and other behavioral data.

[0193] "Emotional analysis" is a process that evaluates a user's emotional state based on voice and behavioral data.

[0194] "Empathic response" refers to a response that aligns with and shows understanding of the user's emotions.

[0195] "Speech synthesis" is a technology that converts generated text into a speech format.

[0196] "Empathic and appropriate responses" refer to communication content that is relevant to the situation and emotions, generated based on the results of emotion analysis.

[0197] The system of the present invention is implemented as a care support tool using an emotion-empathetic AI agent. This system aims to improve the quality of care by understanding the emotions of the elderly and automatically generating empathetic and appropriate dialogue.

[0198] The server first collects the user's voice and converts the analog audio signal into digital data. The audio data is noise-filtered, and features are extracted using Mel-frequency cepstrum coefficients. In addition, user behavior data is collected. This data is sent to the emotion engine, where a deep learning-based machine learning model precisely analyzes the user's emotions.

[0199] Based on the analysis results, the server uses natural language processing technology to generate a user-centric response. This response is synthesized as natural speech by a speech synthesis engine and played back to the user through the speaker. Providing an appropriate response that is relevant to the user's current situation improves the user experience.

[0200] For example, if a user says, "I'm feeling a little down today," the emotion engine can identify the corresponding emotion, and the server can generate a response such as, "Is there anything I can do to help?" and deliver it via voice. Therefore, accuracy in emotion recognition and the ability to have natural conversations are considered important.

[0201] An example of a prompt to the generation AI model is, "Based on the user's voice, please understand their current emotions and come up with appropriate words of encouragement." This prompt enables the system to generate and provide emotionally resonant and appropriate dialogue.

[0202] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0203] Step 1:

[0204] The device collects the user's voice in real time through the microphone. The collected analog audio signal is converted into digital data. This digital data is then filtered to remove unwanted sounds and used as input data.

[0205] Step 2:

[0206] The device extracts speech features from noise-filtered digital data using Mel-frequency cepstrum coefficients (MFCCs). The output is feature data that shows the frequency pattern of the speech.

[0207] Step 3:

[0208] The device collects user behavior data as additional input. This behavior data includes gestures and actions, and is also recorded in digital format.

[0209] Step 4:

[0210] The device sends voice feature data and behavioral data to the server. The server inputs this data into an emotion engine and analyzes the user's emotions using a generative AI model. The output of this step is the identified emotion label.

[0211] Step 5:

[0212] The server uses natural language processing technology to generate empathetic responses tailored to the user based on the analyzed emotion labels. The responses are output in text format, and their content is relevant to the user's situation and emotions.

[0213] Step 6:

[0214] The server inputs the generated text response into a speech synthesis engine, which converts it into natural and expressive speech data. The output is the synthesized speech file.

[0215] Step 7:

[0216] The device plays synthesized speech to the user through its speaker. This playback process allows the user to receive empathetic messages at appropriate times.

[0217] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0218] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0219] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0220] [Second Embodiment]

[0221] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0222] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0223] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0224] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0225] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0226] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0227] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0228] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0229] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0230] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0231] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0232] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0233] The empathetic AI agent system according to the present invention collects voice from the user in real time and emotionally analyzes the voice to provide the user with a more appropriate and emotionally rich response. An embodiment of this system is shown below.

[0234] First, the terminal collects the user's voice through a voice input device. This voice is converted into digital data in real time. Next, the terminal extracts features from this digital data and sends them to the server as samples. In particular, feature extraction is performed from the voice data using Mel-frequency cepstrum coefficients (MFCCs).

[0235] The server uses the received audio feature data to run machine learning algorithms and analyze the user's emotions. A pre-trained deep learning model is used for this, and emotion labels are assigned to the analysis results. Based on these emotion labels, the server utilizes natural language processing techniques to generate an appropriate response in text format.

[0236] The generated response text is again converted into speech data by the server using speech synthesis technology. This speech data is synthesized with attention to emotional nuances and is ultimately sent to the terminal.

[0237] The device plays the received synthesized speech data to the user through its speaker. This allows the user to receive responses that are natural and relatable, tailored to their own emotions.

[0238] As a concrete example, consider a scenario where a user says, "I'm tired today." In this case, the voice is collected by the device, and emotion analysis identifies the emotion as "fatigue." The response generation process creates a message such as, "You've worked hard. Please get plenty of rest," and the voice is synthesized and played back to the user. As a result, the user receives a response that reflects their own emotions, making the interaction with the system more natural. Thus, this invention realizes dialogue with an AI agent that matches the user's emotions, thereby improving the quality of life.

[0239] The following describes the processing flow.

[0240] Step 1:

[0241] The device collects the user's voice through the microphone. The collected voice is converted into a digital signal, and noise is filtered out to produce clean audio data.

[0242] Step 2:

[0243] The terminal divides the digital signal into short segments and extracts speech features from each segment. Here, Mel-frequency cepstrum coefficients (MFCCs) are used to calculate the speech features.

[0244] Step 3:

[0245] The terminal sends the extracted feature data to the server. This transmission may involve data compression to optimize network efficiency.

[0246] Step 4:

[0247] The server uses the received audio feature data to perform sentiment analysis using a machine learning model. The analysis generates labels indicating emotions and identifies the user's emotional state.

[0248] Step 5:

[0249] The server generates response text that matches the original user statement based on the sentiment analysis results. Natural language processing techniques are used to select wording that reflects the user's emotions.

[0250] Step 6:

[0251] The server converts the generated response text into speech data using a speech synthesis engine. During this process, it adjusts intonation, pitch, and speed to express emotional nuances.

[0252] Step 7:

[0253] The server sends the synthesized audio data to the terminal. This transmission also involves data compression as needed.

[0254] Step 8:

[0255] The device plays the received audio data through its speaker. In this way, it provides the user with an emotionally resonant voice response.

[0256] (Example 1)

[0257] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0258] Traditional interaction technologies face the challenge of generating empathetic responses to user emotions in real time. Furthermore, if the generated responses do not adequately reflect the nuances of those emotions, communication with the user tends to become strained.

[0259] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0260] In this invention, the server includes means for acquiring the user's voice and converting it into an electrical signal, means for reading features from the electrical signal, and means for evaluating emotions using the read features. This enables the server to appropriately evaluate the user's emotions in real time, quickly generate empathetic responses based on those evaluations, and facilitate natural communication with the user.

[0261] "Voice" refers to the audio signals emitted by the user, which are collected as acoustic information.

[0262] An "electrical signal" is a format used for information processing, which is the conversion of sound into digital data.

[0263] A "feature" is a specific element of data extracted from an audio signal that contains information necessary for sentiment evaluation.

[0264] "Emotion" refers to the psychological state a user is in, and is identified through their vocal characteristics.

[0265] "Evaluation" is the process of determining the user's emotions based on features extracted from their voice.

[0266] "Response" refers to an audio or text message generated based on the user's sentiment evaluation.

[0267] "Synthesis" refers to the process of generating speech from text or data, and is a technology for providing voice responses to users.

[0268] "Output" refers to playing the generated audio through a device such as a speaker.

[0269] The emotion-empathizing AI agent system according to this invention uses multiple hardware and software components to provide natural interaction based on the user's emotions.

[0270] First, the terminal uses a smartphone or computer equipped with a microphone as an audio input device. The user's voice is picked up by the microphone as an analog signal and converted into an electrical signal on the spot. The electrical signal is sampled over a specific time interval through signal processing technology, and features are extracted as digital data. For this process, Mel-frequency cepstrum coefficients (MFCCs) are used as a method for frequency conversion.

[0271] Next, the server receives feature data sent from the terminal and uses a pre-trained learning algorithm to evaluate the user's emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used for emotion evaluation. As a result of the evaluation, emotion labels such as "joy," "sadness," and "anger" are assigned.

[0272] Based on the received emotion label, the server uses natural language processing techniques to generate an empathetic response. This response generation utilizes a generative AI model, which generates appropriate text in response to prompt text. For example, the prompt text "What should I say if the user's emotion is 'fatigue'?" is input to the model.

[0273] The generated text responses are converted into audio data using speech synthesis technology. Technologies such as WaveNet are used for speech synthesis to achieve natural and fluent expression. The audio data is then sent back to the terminal.

[0274] Ultimately, the device outputs synthesized speech data to the user through its speaker. This allows the user to naturally receive empathetic responses from the system, resulting in a more human-like interface.

[0275] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0276] Step 1:

[0277] The user emits sound through the microphone. This sound is input to the device and captured as an analog signal. Next, the device converts the analog signal into an electrical signal. The resulting electrical signal becomes the input for the audio data.

[0278] Step 2:

[0279] The terminal converts this electrical signal into digital audio data and samples it over a specific time interval. From the sampled digital data, it performs frequency conversion to extract Mel-frequency cepstrum coefficients (MFCCs). The MFCCs become the output feature data, representing important characteristics of the audio as numerical data.

[0280] Step 3:

[0281] The terminal sends the extracted MFCC data to the server as sample data. The server receives this data and uses it as input data for sentiment analysis. This transmitted feature data is the input for the next process.

[0282] Step 4:

[0283] The server uses the received feature data to execute a deep learning model to evaluate emotions. Here, convolutional neural networks (CNNs) or recurrent neural networks (RNNs) are used. As a result, emotion labels are output. For example, emotions such as "joy", "anger", and "sadness" are analyzed as emotion labels.

[0284] Step 5:

[0285] Based on the emotion labels, the server uses natural language processing technology to generate empathetic response texts. In this process, a generative AI model is utilized, and a prompt sentence is input into the model to output appropriate text. For example, a prompt such as "When the user's emotion is 'fatigue', what words should be used to talk to them?" is used.

[0286] Step 6:

[0287] The server utilizes speech synthesis technology to convert the generated response text into speech. The generated audio data is synthesized using a model such as WaveNet, and an output as speech that conforms to the emotion is obtained.

[0288] Step 7:

[0289] The terminal receives the synthesized audio data transmitted from the server and plays it through a speaker. Through this output audio, the user can receive a response that empathizes with their emotion.

[0290] (Application Example 1)

[0291] Next, Application Example 1 will be described. In the following description, the data processing device 12 is referred to as the "server", and the smart glasses 214 are referred to as the "terminal".

[0292] Communicating with the elderly requires understanding their emotions appropriately and providing empathetic responses. However, conventional techniques are insufficient in adequately addressing the unique emotional states of the elderly. Therefore, there is a need for more effective communication methods to improve the quality of life for the elderly.

[0293] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0294] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, and means for generating responses that are empathetic to the emotions of elderly people based on the results of emotion analysis. This makes it possible to understand the unique emotional state of elderly people and to communicate appropriately with them.

[0295] "Sound" refers to the audio signal received from the user, which is the input data that the system uses to convert it into digital data and analyze it.

[0296] "Digital data" refers to data obtained by converting audio into a format that can be processed on a computer, and it serves as fundamental information used for feature extraction and sentiment analysis.

[0297] "Feature extraction" is the process of extracting specific information from audio data and deriving metrics necessary for data analysis. It is a crucial process that affects the accuracy of the analysis.

[0298] "Emotional analysis" is a process that analyzes the emotions contained in a user's utterances based on feature-extracted data and determines a specific emotional state.

[0299] "Response generation" is the process of creating appropriate text responses based on the results of sentiment analysis, enabling communication that is sensitive to the user's emotions.

[0300] "Speech synthesis" is the process of converting text-generated responses back into speech format, providing users with natural-sounding responses.

[0301] "Empathizing with the emotions of the elderly" means understanding the unique emotional states that elderly people exhibit, responding appropriately and reassuringly, and aiming to achieve comfortable communication.

[0302] To implement this invention, it is necessary to use a terminal that functions as a user interface and a server that performs voice data analysis. The terminal is equipped with a voice input device that collects the user's voice in real time and converts it into digital data.

[0303] The digital data converted by the terminal is sent to a server for feature extraction. The server uses software such as Python or LibROSA to extract Mel-frequency cepstrum coefficients. This allows the necessary features for sentiment analysis to be obtained from the audio data.

[0304] Sentiment analysis is performed on a server using a pre-trained deep learning model. This process utilizes machine learning frameworks such as TensorFlow to assign labels indicating specific emotions to audio data. Then, Hugging Face's Transformers are used to generate natural-sounding responses based on the analysis results.

[0305] The generated responses are converted into audio data using speech synthesis software such as Google Text-to-Speech. The converted audio data is sent to the terminal and played back to the user. This entire process enables the system to communicate in a way that is sensitive to the emotions of elderly people.

[0306] As a specific example, when a user says, "My family rarely comes," the server can generate a response that empathizes with the emotion, such as "Your family is very important. Shall we think about what you would like to talk about next time they visit?" An example of a prompt sentence could be, "Please generate a response that reflects the following emotion: sadness - appropriate comfort when the conversation partner is sad."

[0307] The flow of the specific process in Application Example 1 will be described using FIG. 12.

[0308] Step 1:

[0309] The terminal collects the user's voice using a voice input device. The input is an analog voice signal, which is converted into digital data. The converted digital data is ready to be transmitted to the server.

[0310] Step 2:

[0311] The server processes the digital voice data received from the terminal and performs feature extraction. Specifically, the Mel-frequency cepstral coefficients are calculated using the LibROSA library. The input is digital voice data, and the output is feature data. As a result, the characteristics of the voice are quantified and can be used for emotion analysis.

[0312] Step 3:

[0313] The server performs emotion analysis based on the feature data. A deep learning model (using TensorFlow) generates an emotion label. The input is feature data, and the output is an emotion label. This label indicates the user's emotional state.

[0314] Step 4:

[0315] The server generates a response using natural language processing techniques based on the generated emotion labels. A text-based response is created using Hugging Face's Transformers. The input is the emotion label, and the output is the text response to be returned to the user.

[0316] Step 5:

[0317] The server converts text responses into speech data. It uses Google Text-to-Speech to synthesize speech from text. The input is the text response, and the output is synthesized speech data. This generates natural-sounding speech responses.

[0318] Step 6:

[0319] The terminal receives synthesized speech data from the server and plays it back to the user through the speaker. The input is synthesized speech data, and the output is the voice the user hears. This enables communication that is sensitive to the emotions of elderly people.

[0320] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0321] The emotion-empathizing AI agent system according to the present invention provides a new interactive system that combines an emotion engine to highly recognize the user's emotions.

[0322] First, the device uses its voice input function to collect the user's voice and converts the analog audio signal into digital data. In this process, the audio data is processed through noise filtering to make it clearer.

[0323] Next, the device extracts voice features from this digital data using Mel-frequency cepstrum coefficients (MFCCs), and in addition, collects user behavior data (e.g., gestures and behaviors), which it then sends to the emotion engine.

[0324] The server drives an emotion engine, integrating voice and behavioral data to analyze the user's emotions. A deep learning-based machine learning model is used for emotion recognition, which generates a variety of emotion labels. The emotion engine precisely identifies multiple emotional states, and the analysis results are stored on the server.

[0325] Based on the analysis results, the server utilizes natural language processing technology to generate responses that resonate with the user's emotions. The content and tone of the responses are dynamically adjusted based on the emotion recognition results. The generated response text is then converted into natural, emotionally expressive speech data by a speech synthesis engine.

[0326] Ultimately, the device plays synthesized speech to the user. This playback is done through the speaker, improving the user experience by providing responses appropriate to the situation the user is facing.

[0327] For example, if a user says "I'm sad today," the emotion engine immediately recognizes the emotion of "sadness." The server then generates a response of encouragement such as "You've had a tough day, please let me know if there's anything I can do to help," which is then played back via speech synthesis, providing the user with an empathetic interaction. In this way, the present invention provides a system that realizes a rich emotional experience for the user.

[0328] The following describes the processing flow.

[0329] Step 1:

[0330] The device collects the user's voice through a microphone. This voice is acquired as an analog signal and immediately converted into digital data. Furthermore, the voice data is processed using noise filtering to remove unwanted background noise and prepare it for voice analysis.

[0331] Step 2:

[0332] The device extracts features from digital audio data using Mel-frequency cepstrum coefficients (MFCCs). These features are numerical data that represent the characteristics of the voice and are used as basic data for speech recognition and sentiment analysis. The device also simultaneously collects user behavior data (e.g., gesture recognition via camera).

[0333] Step 3:

[0334] The terminal combines the extracted voice data and behavioral data and sends it to the server. The data is compressed during transmission to reduce network load.

[0335] Step 4:

[0336] The server analyzes the received data using an emotion engine. The emotion engine integrates voice and behavioral data using machine learning models to identify the user's emotional state. The identified emotions are categorized into specific labels, and the results are processed on the server.

[0337] Step 5:

[0338] The server applies a natural language processing algorithm based on the sentiment analysis results to generate the optimal response for the user. This response is designed with content and tone that resonates with the user's current emotions.

[0339] Step 6:

[0340] The server uses the generated response text to initiate the speech synthesis process. This process synthesizes speech that reflects emotional expression from the generated text, constructing the final audio file.

[0341] Step 7:

[0342] The server sends the synthesized audio to the terminal. The terminal also decodes the data as needed and prepares it to be delivered to the user in the optimal state.

[0343] Step 8:

[0344] The device plays the received synthesized speech through its speaker. This allows the user to hear responses in a voice and expression that reflects their own emotions, enabling them to enjoy a natural dialogue with the system.

[0345] (Example 2)

[0346] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0347] Conventional voice dialogue systems face the challenge of not being able to appropriately recognize user emotions and provide empathetic responses. Furthermore, it is necessary to perform more accurate emotion analysis by taking into account not only voice information but also behavioral information.

[0348] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0349] In this invention, the server includes means for acquiring user voice and converting it into digital information, means for extracting characteristics from the digital information, and means for acquiring user behavior and using it for emotion analysis. This enables integrated analysis of emotions from voice and behavior, and the generation of responses that are tailored to the user.

[0350] "User voice" refers to the human voice signal input into the system.

[0351] "Digital information" refers to data obtained by converting analog signals into a digital format.

[0352] "Characteristics" refer to distinctive elements or patterns contained in voice or behavioral data.

[0353] "Emotions" refer to information that indicates the user's psychological state.

[0354] A "response" is a message or action that a system returns to the user.

[0355] "Behavior" refers to information obtained from the user's gestures and actions.

[0356] The "frequency coefficient" is an index used to analyze the characteristics of audio data.

[0357] A "machine learning algorithm" is a technology that learns patterns from data and uses them to make predictions and classifications.

[0358] "Natural language processing technology" is a technology that enables computers to understand and generate human language.

[0359] In this invention, the user first inputs their voice into the system using a terminal equipped with an audio input device. The terminal is equipped with a high-performance microphone that can accurately capture the voice. The audio signal is converted from analog to digital, and noise is removed using digital signal processing technology. This results in clear audio data.

[0360] Next, the device uses signal processing software to extract speech characteristics by calculating Mel-frequency cepstrum coefficients (MFCCs) from the digitized speech data. It also captures user gestures and movements through cameras and motion sensors, collecting this as behavioral data. This consistent data, combined with speech, forms the basis for sentiment analysis.

[0361] The server uses a generative AI model powered by deep learning to evaluate integrated speech and behavioral data through sentiment analysis. This model is trained on a large amount of data and accurately identifies human emotions as multiple emotion labels. The identified emotions are stored on the server and updated as needed.

[0362] The server then uses natural language processing technology to generate a response that matches the user's emotional state. The generated response is adjusted in content and tone to be more natural and empathetic to the user's feelings. For example, if the user says, "I'm sad today," the system recognizes the emotion of "sadness" and generates a response such as, "You've had a tough day, please let me know if there's anything I can do to help."

[0363] The generated text is converted into expressive speech through a speech synthesis engine. The device then plays this synthesized speech through its speaker, delivering it directly to the user. This process allows the user to have an experience that feels like interacting with a human, rather than a robotic one.

[0364] Example of a prompt

[0365] "Today was very busy, and I'm a little tired. Could you tell me how to relax?"

[0366] "I'm feeling a little anxious about the new project. Could you give me some advice?"

[0367] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0368] Step 1:

[0369] The user inputs their voice into the device via the microphone. The device receives this voice as an analog signal and converts it to a digital signal using an analog-to-digital converter. This converted digital audio data is then subjected to a noise reduction filter to reduce background noise and obtain a clear result. This then becomes the input for the next step.

[0370] Step 2:

[0371] The device uses the cleared digital audio data to extract Mel-frequency cepstrum coefficients (MFCCs) using a signal processing algorithm. This analyzes the characteristics of the audio, which are then output as feature data. In addition, the camera and motion sensors capture the user's gestures and movements, which are stored as behavioral data. This data is then transmitted to the emotion analysis engine.

[0372] Step 3:

[0373] The server receives voice feature data and behavioral data sent from the terminal. Using a generative AI model, this data is integrated and sentiment analysis is performed. This analysis classifies the data into numerous sentiment labels, allowing for an understanding of the user's emotional state. This is crucial input for the next step.

[0374] Step 4:

[0375] The server uses natural language processing technology to generate user-specific responses based on the sentiment analysis results. The content and tone of the responses are adjusted according to the analyzed sentiment labels. The generated responses are output in text format.

[0376] Step 5:

[0377] The terminal receives generated text sent from the server and converts it into speech data using a speech synthesis engine. Speech synthesis produces speech with natural and emotional intonation, which becomes the final output.

[0378] Step 6:

[0379] The device plays synthesized speech to the user via its speaker. At this point, the user receives empathetic responses from the system in voice, completing the conversational experience.

[0380] (Application Example 2)

[0381] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0382] In elderly care settings, it is crucial to accurately understand changes in emotions and conditions, and to automatically provide empathetic dialogue and instructions accordingly. A lack of emotional understanding can lead to a decline in the satisfaction and quality of life of elderly individuals. A system is needed to address this situation while simultaneously reducing the burden on caregivers.

[0383] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0384] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, means for performing sentiment analysis on the extracted feature data and behavioral data, means for generating empathetic responses based on the analysis results, and means for synthesizing the generated responses as voice and making suggestions based on a specific context. This makes it possible to accurately grasp the emotional state of the person receiving care and provide empathetic and appropriate responses.

[0385] A "user" refers to a person who uses a system and interacts with it.

[0386] "Speech" refers to a waveform signal containing language emitted by the user.

[0387] "Digital data" is a data format that quantifies analog audio signals.

[0388] "Feature extraction" is the process of extracting meaningful patterns or information from digital data.

[0389] "Behavioral data" refers to information about a user's actions, gaze, gestures, and other behavioral data.

[0390] "Emotional analysis" is a process that evaluates a user's emotional state based on voice and behavioral data.

[0391] "Empathic response" refers to a response that aligns with and shows understanding of the user's emotions.

[0392] "Speech synthesis" is a technology that converts generated text into a speech format.

[0393] "Empathic and appropriate responses" refer to communication content that is relevant to the situation and emotions, generated based on the results of emotion analysis.

[0394] The system of the present invention is implemented as a care support tool using an emotion-empathetic AI agent. This system aims to improve the quality of care by understanding the emotions of the elderly and automatically generating empathetic and appropriate dialogue.

[0395] The server first collects the user's voice and converts the analog audio signal into digital data. The audio data is noise-filtered, and features are extracted using Mel-frequency cepstrum coefficients. In addition, user behavior data is collected. This data is sent to the emotion engine, where a deep learning-based machine learning model precisely analyzes the user's emotions.

[0396] Based on the analysis results, the server uses natural language processing technology to generate a user-centric response. This response is synthesized as natural speech by a speech synthesis engine and played back to the user through the speaker. Providing an appropriate response that is relevant to the user's current situation improves the user experience.

[0397] For example, if a user says, "I'm feeling a little down today," the emotion engine can identify the corresponding emotion, and the server can generate a response such as, "Is there anything I can do to help?" and deliver it via voice. Therefore, accuracy in emotion recognition and the ability to have natural conversations are considered important.

[0398] An example of a prompt to the generation AI model is, "Based on the user's voice, please understand their current emotions and come up with appropriate words of encouragement." This prompt enables the system to generate and provide emotionally resonant and appropriate dialogue.

[0399] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0400] Step 1:

[0401] The device collects the user's voice in real time through the microphone. The collected analog audio signal is converted into digital data. This digital data is then filtered to remove unwanted sounds and used as input data.

[0402] Step 2:

[0403] The device extracts speech features from noise-filtered digital data using Mel-frequency cepstrum coefficients (MFCCs). The output is feature data that shows the frequency pattern of the speech.

[0404] Step 3:

[0405] The device collects user behavior data as additional input. This behavior data includes gestures and actions, and is also recorded in digital format.

[0406] Step 4:

[0407] The device sends voice feature data and behavioral data to the server. The server inputs this data into an emotion engine and analyzes the user's emotions using a generative AI model. The output of this step is the identified emotion label.

[0408] Step 5:

[0409] The server uses natural language processing technology to generate empathetic responses tailored to the user based on the analyzed emotion labels. The responses are output in text format, and their content is relevant to the user's situation and emotions.

[0410] Step 6:

[0411] The server inputs the generated text response into a speech synthesis engine, which converts it into natural and expressive speech data. The output is the synthesized speech file.

[0412] Step 7:

[0413] The device plays synthesized speech to the user through its speaker. This playback process allows the user to receive empathetic messages at appropriate times.

[0414] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0415] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0416] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0417] [Third Embodiment]

[0418] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0419] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0420] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0421] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0422] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0423] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0424] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0425] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0426] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0427] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0428] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0429] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0430] The empathetic AI agent system according to the present invention collects voice from the user in real time and emotionally analyzes the voice to provide the user with a more appropriate and emotionally rich response. An embodiment of this system is shown below.

[0431] First, the terminal collects the user's voice through a voice input device. This voice is converted into digital data in real time. Next, the terminal extracts features from this digital data and sends them to the server as samples. In particular, feature extraction is performed from the voice data using Mel-frequency cepstrum coefficients (MFCCs).

[0432] The server uses the received audio feature data to run machine learning algorithms and analyze the user's emotions. A pre-trained deep learning model is used for this, and emotion labels are assigned to the analysis results. Based on these emotion labels, the server utilizes natural language processing techniques to generate an appropriate response in text format.

[0433] The generated response text is again converted into speech data by the server using speech synthesis technology. This speech data is synthesized with attention to emotional nuances and is ultimately sent to the terminal.

[0434] The device plays the received synthesized speech data to the user through its speaker. This allows the user to receive responses that are natural and relatable, tailored to their own emotions.

[0435] As a concrete example, consider a scenario where a user says, "I'm tired today." In this case, the voice is collected by the device, and emotion analysis identifies the emotion as "fatigue." The response generation process creates a message such as, "You've worked hard. Please get plenty of rest," and the voice is synthesized and played back to the user. As a result, the user receives a response that reflects their own emotions, making the interaction with the system more natural. Thus, this invention realizes dialogue with an AI agent that matches the user's emotions, thereby improving the quality of life.

[0436] The following describes the processing flow.

[0437] Step 1:

[0438] The device collects the user's voice through the microphone. The collected voice is converted into a digital signal, and noise is filtered out to produce clean audio data.

[0439] Step 2:

[0440] The terminal divides the digital signal into short segments and extracts speech features from each segment. Here, Mel-frequency cepstrum coefficients (MFCCs) are used to calculate the speech features.

[0441] Step 3:

[0442] The terminal sends the extracted feature data to the server. This transmission may involve data compression to optimize network efficiency.

[0443] Step 4:

[0444] The server uses the received audio feature data to perform sentiment analysis using a machine learning model. The analysis generates labels indicating emotions and identifies the user's emotional state.

[0445] Step 5:

[0446] The server generates response text that matches the original user statement based on the sentiment analysis results. Natural language processing techniques are used to select wording that reflects the user's emotions.

[0447] Step 6:

[0448] The server converts the generated response text into speech data using a speech synthesis engine. During this process, it adjusts intonation, pitch, and speed to express emotional nuances.

[0449] Step 7:

[0450] The server sends the synthesized audio data to the terminal. This transmission also involves data compression as needed.

[0451] Step 8:

[0452] The device plays the received audio data through its speaker. In this way, it provides the user with an emotionally resonant voice response.

[0453] (Example 1)

[0454] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0455] Traditional interaction technologies face the challenge of generating empathetic responses to user emotions in real time. Furthermore, if the generated responses do not adequately reflect the nuances of those emotions, communication with the user tends to become strained.

[0456] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0457] In this invention, the server includes means for acquiring the user's voice and converting it into an electrical signal, means for reading features from the electrical signal, and means for evaluating emotions using the read features. This enables the server to appropriately evaluate the user's emotions in real time, quickly generate empathetic responses based on those evaluations, and facilitate natural communication with the user.

[0458] "Voice" refers to the audio signals emitted by the user, which are collected as acoustic information.

[0459] An "electrical signal" is a format used for information processing, which is the conversion of sound into digital data.

[0460] A "feature" is a specific element of data extracted from an audio signal that contains information necessary for sentiment evaluation.

[0461] "Emotion" refers to the psychological state a user is in, and is identified through their vocal characteristics.

[0462] "Evaluation" is the process of determining the user's emotions based on features extracted from their voice.

[0463] "Response" refers to an audio or text message generated based on the user's sentiment evaluation.

[0464] "Synthesis" refers to the process of generating speech from text or data, and is a technology for providing voice responses to users.

[0465] "Output" refers to playing the generated audio through a device such as a speaker.

[0466] The emotion-empathizing AI agent system according to this invention uses multiple hardware and software components to provide natural interaction based on the user's emotions.

[0467] First, the terminal uses a smartphone or computer equipped with a microphone as an audio input device. The user's voice is picked up by the microphone as an analog signal and converted into an electrical signal on the spot. The electrical signal is sampled over a specific time interval through signal processing technology, and features are extracted as digital data. For this process, Mel-frequency cepstrum coefficients (MFCCs) are used as a method for frequency conversion.

[0468] Next, the server receives feature data sent from the terminal and uses a pre-trained learning algorithm to evaluate the user's emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used for emotion evaluation. As a result of the evaluation, emotion labels such as "joy," "sadness," and "anger" are assigned.

[0469] Based on the received emotion label, the server uses natural language processing techniques to generate an empathetic response. This response generation utilizes a generative AI model, which generates appropriate text in response to prompt text. For example, the prompt text "What should I say if the user's emotion is 'fatigue'?" is input to the model.

[0470] The generated text responses are converted into audio data using speech synthesis technology. Technologies such as WaveNet are used for speech synthesis to achieve natural and fluent expression. The audio data is then sent back to the terminal.

[0471] Ultimately, the device outputs synthesized speech data to the user through its speaker. This allows the user to naturally receive empathetic responses from the system, resulting in a more human-like interface.

[0472] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0473] Step 1:

[0474] The user emits sound through the microphone. This sound is input to the device and captured as an analog signal. Next, the device converts the analog signal into an electrical signal. The resulting electrical signal becomes the input for the audio data.

[0475] Step 2:

[0476] The terminal converts this electrical signal into digital audio data and samples it over a specific time interval. From the sampled digital data, it performs frequency conversion to extract Mel-frequency cepstrum coefficients (MFCCs). The MFCCs become the output feature data, representing important characteristics of the audio as numerical data.

[0477] Step 3:

[0478] The terminal sends the extracted MFCC data to the server as sample data. The server receives this data and uses it as input data for sentiment analysis. This transmitted feature data is the input for the next process.

[0479] Step 4:

[0480] The server uses the received feature data to run a deep learning model and evaluate emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used here. This outputs emotion labels. For example, emotions such as "joy," "anger," and "sadness" are analyzed.

[0481] Step 5:

[0482] The server generates empathetic response text using natural language processing techniques based on emotion labels. This process utilizes a generative AI model, which is input with prompts to output appropriate text. For example, a prompt might ask, "What should I say if the user's emotion is 'fatigue'?"

[0483] Step 6:

[0484] The server uses speech synthesis technology to convert the generated response text into speech. The generated speech data is synthesized using models such as WaveNet to produce output that reflects emotions.

[0485] Step 7:

[0486] The device receives synthesized speech data sent from the server and plays it through the speaker. Through this outputted speech, the user can receive emotionally resonant responses.

[0487] (Application Example 1)

[0488] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0489] Communicating with the elderly requires understanding their emotions appropriately and providing empathetic responses. However, conventional techniques are insufficient in adequately addressing the unique emotional states of the elderly. Therefore, there is a need for more effective communication methods to improve the quality of life for the elderly.

[0490] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0491] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, and means for generating responses that are empathetic to the emotions of elderly people based on the results of emotion analysis. This makes it possible to understand the unique emotional state of elderly people and to communicate appropriately with them.

[0492] "Sound" refers to the audio signal received from the user, which is the input data that the system uses to convert it into digital data and analyze it.

[0493] "Digital data" refers to data obtained by converting audio into a format that can be processed on a computer, and it serves as fundamental information used for feature extraction and sentiment analysis.

[0494] "Feature extraction" is the process of extracting specific information from audio data and deriving metrics necessary for data analysis. It is a crucial process that affects the accuracy of the analysis.

[0495] "Emotional analysis" is a process that analyzes the emotions contained in a user's utterances based on feature-extracted data and determines a specific emotional state.

[0496] "Response generation" is the process of creating appropriate text responses based on the results of sentiment analysis, enabling communication that is sensitive to the user's emotions.

[0497] "Speech synthesis" is the process of converting text-generated responses back into speech format, providing users with natural-sounding responses.

[0498] "Empathizing with the emotions of the elderly" means understanding the unique emotional states that elderly people exhibit, responding appropriately and reassuringly, and aiming to achieve comfortable communication.

[0499] To implement this invention, it is necessary to use a terminal that functions as a user interface and a server that performs voice data analysis. The terminal is equipped with a voice input device that collects the user's voice in real time and converts it into digital data.

[0500] The digital data converted by the terminal is sent to a server for feature extraction. The server uses software such as Python or LibROSA to extract Mel-frequency cepstrum coefficients. This allows the necessary features for sentiment analysis to be obtained from the audio data.

[0501] Sentiment analysis is performed on a server using a pre-trained deep learning model. This process utilizes machine learning frameworks such as TensorFlow to assign labels indicating specific emotions to audio data. Then, Hugging Face's Transformers are used to generate natural-sounding responses based on the analysis results.

[0502] The generated responses are converted into audio data using speech synthesis software such as Google Text-to-Speech. The converted audio data is sent to the terminal and played back to the user. This entire process enables the system to communicate in a way that is sensitive to the emotions of elderly people.

[0503] For example, if a user says, "My family isn't coming," the server can generate an empathetic response such as, "They are important family members, aren't they? Shall I think about what you'd like to talk about next time they visit?" An example of a prompt might be, "Generate a response that reflects the following emotion: Sadness - Appropriate comfort when the person you're talking to is sad."

[0504] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0505] Step 1:

[0506] The terminal collects the user's voice using an audio input device. The input is an analog audio signal, which is then converted into digital data. The converted digital data is then ready for transmission to the server.

[0507] Step 2:

[0508] The server processes the digital audio data received from the terminal and performs feature extraction. Specifically, it uses the LibROSA library to calculate Mel-frequency cepstrum coefficients. The input is digital audio data, and the output is feature data. This allows the characteristics of the audio to be quantified and used for sentiment analysis.

[0509] Step 3:

[0510] The server performs sentiment analysis based on feature data. A deep learning model (using TensorFlow) generates sentiment labels. The input is feature data, and the output is sentiment labels. These labels indicate the user's emotional state.

[0511] Step 4:

[0512] The server generates a response using natural language processing techniques based on the generated emotion labels. A text-based response is created using Hugging Face's Transformers. The input is the emotion label, and the output is the text response to be returned to the user.

[0513] Step 5:

[0514] The server converts text responses into speech data. It uses Google Text-to-Speech to synthesize speech from text. The input is the text response, and the output is synthesized speech data. This generates natural-sounding speech responses.

[0515] Step 6:

[0516] The terminal receives synthesized speech data from the server and plays it back to the user through the speaker. The input is synthesized speech data, and the output is the voice the user hears. This enables communication that is sensitive to the emotions of elderly people.

[0517] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0518] The emotion-empathizing AI agent system according to the present invention provides a new interactive system that combines an emotion engine to highly recognize the user's emotions.

[0519] First, the device uses its voice input function to collect the user's voice and converts the analog audio signal into digital data. In this process, the audio data is processed through noise filtering to make it clearer.

[0520] Next, the device extracts voice features from this digital data using Mel-frequency cepstrum coefficients (MFCCs), and in addition, collects user behavior data (e.g., gestures and behaviors), which it then sends to the emotion engine.

[0521] The server drives an emotion engine, integrating voice and behavioral data to analyze the user's emotions. A deep learning-based machine learning model is used for emotion recognition, which generates a variety of emotion labels. The emotion engine precisely identifies multiple emotional states, and the analysis results are stored on the server.

[0522] Based on the analysis results, the server utilizes natural language processing technology to generate responses that resonate with the user's emotions. The content and tone of the responses are dynamically adjusted based on the emotion recognition results. The generated response text is then converted into natural, emotionally expressive speech data by a speech synthesis engine.

[0523] Ultimately, the device plays synthesized speech to the user. This playback is done through the speaker, improving the user experience by providing responses appropriate to the situation the user is facing.

[0524] For example, if a user says "I'm sad today," the emotion engine immediately recognizes the emotion of "sadness." The server then generates a response of encouragement such as "You've had a tough day, please let me know if there's anything I can do to help," which is then played back via speech synthesis, providing the user with an empathetic interaction. In this way, the present invention provides a system that realizes a rich emotional experience for the user.

[0525] The following describes the processing flow.

[0526] Step 1:

[0527] The device collects the user's voice through a microphone. This voice is acquired as an analog signal and immediately converted into digital data. Furthermore, the voice data is processed using noise filtering to remove unwanted background noise and prepare it for voice analysis.

[0528] Step 2:

[0529] The device extracts features from digital audio data using Mel-frequency cepstrum coefficients (MFCCs). These features are numerical data that represent the characteristics of the voice and are used as basic data for speech recognition and sentiment analysis. The device also simultaneously collects user behavior data (e.g., gesture recognition via camera).

[0530] Step 3:

[0531] The terminal combines the extracted voice data and behavioral data and sends it to the server. The data is compressed during transmission to reduce network load.

[0532] Step 4:

[0533] The server analyzes the received data using an emotion engine. The emotion engine integrates voice and behavioral data using machine learning models to identify the user's emotional state. The identified emotions are categorized into specific labels, and the results are processed on the server.

[0534] Step 5:

[0535] The server applies a natural language processing algorithm based on the sentiment analysis results to generate the optimal response for the user. This response is designed with content and tone that resonates with the user's current emotions.

[0536] Step 6:

[0537] The server uses the generated response text to initiate the speech synthesis process. This process synthesizes speech that reflects emotional expression from the generated text, constructing the final audio file.

[0538] Step 7:

[0539] The server sends the synthesized audio to the terminal. The terminal also decodes the data as needed and prepares it to be delivered to the user in the optimal state.

[0540] Step 8:

[0541] The device plays the received synthesized speech through its speaker. This allows the user to hear responses in a voice and expression that reflects their own emotions, enabling them to enjoy a natural dialogue with the system.

[0542] (Example 2)

[0543] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0544] Conventional voice dialogue systems face the challenge of not being able to appropriately recognize user emotions and provide empathetic responses. Furthermore, it is necessary to perform more accurate emotion analysis by taking into account not only voice information but also behavioral information.

[0545] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0546] In this invention, the server includes means for acquiring user voice and converting it into digital information, means for extracting characteristics from the digital information, and means for acquiring user behavior and using it for emotion analysis. This enables integrated analysis of emotions from voice and behavior, and the generation of responses that are tailored to the user.

[0547] "User voice" refers to the human voice signal input into the system.

[0548] "Digital information" refers to data obtained by converting analog signals into a digital format.

[0549] "Characteristics" refer to distinctive elements or patterns contained in voice or behavioral data.

[0550] "Emotions" refer to information that indicates the user's psychological state.

[0551] A "response" is a message or action that a system returns to the user.

[0552] "Behavior" refers to information obtained from the user's gestures and actions.

[0553] The "frequency coefficient" is an index used to analyze the characteristics of audio data.

[0554] A "machine learning algorithm" is a technology that learns patterns from data and uses them to make predictions and classifications.

[0555] "Natural language processing technology" is a technology that enables computers to understand and generate human language.

[0556] In this invention, the user first inputs their voice into the system using a terminal equipped with an audio input device. The terminal is equipped with a high-performance microphone that can accurately capture the voice. The audio signal is converted from analog to digital, and noise is removed using digital signal processing technology. This results in clear audio data.

[0557] Next, the device uses signal processing software to extract speech characteristics by calculating Mel-frequency cepstrum coefficients (MFCCs) from the digitized speech data. It also captures user gestures and movements through cameras and motion sensors, collecting this as behavioral data. This consistent data, combined with speech, forms the basis for sentiment analysis.

[0558] The server uses a generative AI model powered by deep learning to evaluate integrated speech and behavioral data through sentiment analysis. This model is trained on a large amount of data and accurately identifies human emotions as multiple emotion labels. The identified emotions are stored on the server and updated as needed.

[0559] The server then uses natural language processing technology to generate a response that matches the user's emotional state. The generated response is adjusted in content and tone to be more natural and empathetic to the user's feelings. For example, if the user says, "I'm sad today," the system recognizes the emotion of "sadness" and generates a response such as, "You've had a tough day, please let me know if there's anything I can do to help."

[0560] The generated text is converted into expressive speech through a speech synthesis engine. The device then plays this synthesized speech through its speaker, delivering it directly to the user. This process allows the user to have an experience that feels like interacting with a human, rather than a robotic one.

[0561] Example of a prompt

[0562] "Today was very busy, and I'm a little tired. Could you tell me how to relax?"

[0563] "I'm feeling a little anxious about the new project. Could you give me some advice?"

[0564] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0565] Step 1:

[0566] The user inputs their voice into the device via the microphone. The device receives this voice as an analog signal and converts it to a digital signal using an analog-to-digital converter. This converted digital audio data is then subjected to a noise reduction filter to reduce background noise and obtain a clear result. This then becomes the input for the next step.

[0567] Step 2:

[0568] The device uses the cleared digital audio data to extract Mel-frequency cepstrum coefficients (MFCCs) using a signal processing algorithm. This analyzes the characteristics of the audio, which are then output as feature data. In addition, the camera and motion sensors capture the user's gestures and movements, which are stored as behavioral data. This data is then transmitted to the emotion analysis engine.

[0569] Step 3:

[0570] The server receives voice feature data and behavioral data sent from the terminal. Using a generative AI model, this data is integrated and sentiment analysis is performed. This analysis classifies the data into numerous sentiment labels, allowing for an understanding of the user's emotional state. This is crucial input for the next step.

[0571] Step 4:

[0572] The server uses natural language processing technology to generate user-specific responses based on the sentiment analysis results. The content and tone of the responses are adjusted according to the analyzed sentiment labels. The generated responses are output in text format.

[0573] Step 5:

[0574] The terminal receives generated text sent from the server and converts it into speech data using a speech synthesis engine. Speech synthesis produces speech with natural and emotional intonation, which becomes the final output.

[0575] Step 6:

[0576] The device plays synthesized speech to the user via its speaker. At this point, the user receives empathetic responses from the system in voice, completing the conversational experience.

[0577] (Application Example 2)

[0578] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0579] In elderly care settings, it is crucial to accurately understand changes in emotions and conditions, and to automatically provide empathetic dialogue and instructions accordingly. A lack of emotional understanding can lead to a decline in the satisfaction and quality of life of elderly individuals. A system is needed to address this situation while simultaneously reducing the burden on caregivers.

[0580] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0581] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, means for performing sentiment analysis on the extracted feature data and behavioral data, means for generating empathetic responses based on the analysis results, and means for synthesizing the generated responses as voice and making suggestions based on a specific context. This makes it possible to accurately grasp the emotional state of the person receiving care and provide empathetic and appropriate responses.

[0582] A "user" refers to a person who uses a system and interacts with it.

[0583] "Speech" refers to a waveform signal containing language emitted by the user.

[0584] "Digital data" is a data format that quantifies analog audio signals.

[0585] "Feature extraction" is the process of extracting meaningful patterns or information from digital data.

[0586] "Behavioral data" refers to information about a user's actions, gaze, gestures, and other behavioral data.

[0587] "Emotional analysis" is a process that evaluates a user's emotional state based on voice and behavioral data.

[0588] "Empathic response" refers to a response that aligns with and shows understanding of the user's emotions.

[0589] "Speech synthesis" is a technology that converts generated text into a speech format.

[0590] "Empathic and appropriate responses" refer to communication content that is relevant to the situation and emotions, generated based on the results of emotion analysis.

[0591] The system of the present invention is implemented as a care support tool using an emotion-empathetic AI agent. This system aims to improve the quality of care by understanding the emotions of the elderly and automatically generating empathetic and appropriate dialogue.

[0592] The server first collects the user's voice and converts the analog audio signal into digital data. The audio data is noise-filtered, and features are extracted using Mel-frequency cepstrum coefficients. In addition, user behavior data is collected. This data is sent to the emotion engine, where a deep learning-based machine learning model precisely analyzes the user's emotions.

[0593] Based on the analysis results, the server uses natural language processing technology to generate a user-centric response. This response is synthesized as natural speech by a speech synthesis engine and played back to the user through the speaker. Providing an appropriate response that is relevant to the user's current situation improves the user experience.

[0594] For example, if a user says, "I'm feeling a little down today," the emotion engine can identify the corresponding emotion, and the server can generate a response such as, "Is there anything I can do to help?" and deliver it via voice. Therefore, accuracy in emotion recognition and the ability to have natural conversations are considered important.

[0595] An example of a prompt to the generation AI model is, "Based on the user's voice, please understand their current emotions and come up with appropriate words of encouragement." This prompt enables the system to generate and provide emotionally resonant and appropriate dialogue.

[0596] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0597] Step 1:

[0598] The device collects the user's voice in real time through the microphone. The collected analog audio signal is converted into digital data. This digital data is then filtered to remove unwanted sounds and used as input data.

[0599] Step 2:

[0600] The device extracts speech features from noise-filtered digital data using Mel-frequency cepstrum coefficients (MFCCs). The output is feature data that shows the frequency pattern of the speech.

[0601] Step 3:

[0602] The device collects user behavior data as additional input. This behavior data includes gestures and actions, and is also recorded in digital format.

[0603] Step 4:

[0604] The device sends voice feature data and behavioral data to the server. The server inputs this data into an emotion engine and analyzes the user's emotions using a generative AI model. The output of this step is the identified emotion label.

[0605] Step 5:

[0606] The server uses natural language processing technology to generate empathetic responses tailored to the user based on the analyzed emotion labels. The responses are output in text format, and their content is relevant to the user's situation and emotions.

[0607] Step 6:

[0608] The server inputs the generated text response into a speech synthesis engine, which converts it into natural and expressive speech data. The output is the synthesized speech file.

[0609] Step 7:

[0610] The device plays synthesized speech to the user through its speaker. This playback process allows the user to receive empathetic messages at appropriate times.

[0611] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0612] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0613] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[0614] [Fourth Embodiment]

[0615] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0616] As shown in Figure 7, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0617] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0618] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[0619] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0620] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0621] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0622] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0623] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0624] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0625] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0626] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0627] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0628] The empathetic AI agent system according to the present invention collects voice from the user in real time and emotionally analyzes the voice to provide the user with a more appropriate and emotionally rich response. An embodiment of this system is shown below.

[0629] First, the terminal collects the user's voice through a voice input device. This voice is converted into digital data in real time. Next, the terminal extracts features from this digital data and sends them to the server as samples. In particular, feature extraction is performed from the voice data using Mel-frequency cepstrum coefficients (MFCCs).

[0630] The server uses the received audio feature data to run machine learning algorithms and analyze the user's emotions. A pre-trained deep learning model is used for this, and emotion labels are assigned to the analysis results. Based on these emotion labels, the server utilizes natural language processing techniques to generate an appropriate response in text format.

[0631] The generated response text is again converted into speech data by the server using speech synthesis technology. This speech data is synthesized with attention to emotional nuances and is ultimately sent to the terminal.

[0632] The device plays the received synthesized speech data to the user through its speaker. This allows the user to receive responses that are natural and relatable, tailored to their own emotions.

[0633] As a concrete example, consider a scenario where a user says, "I'm tired today." In this case, the voice is collected by the device, and emotion analysis identifies the emotion as "fatigue." The response generation process creates a message such as, "You've worked hard. Please get plenty of rest," and the voice is synthesized and played back to the user. As a result, the user receives a response that reflects their own emotions, making the interaction with the system more natural. Thus, this invention realizes dialogue with an AI agent that matches the user's emotions, thereby improving the quality of life.

[0634] The following describes the processing flow.

[0635] Step 1:

[0636] The device collects the user's voice through the microphone. The collected voice is converted into a digital signal, and noise is filtered out to produce clean audio data.

[0637] Step 2:

[0638] The terminal divides the digital signal into short segments and extracts speech features from each segment. Here, Mel-frequency cepstrum coefficients (MFCCs) are used to calculate the speech features.

[0639] Step 3:

[0640] The terminal sends the extracted feature data to the server. This transmission may involve data compression to optimize network efficiency.

[0641] Step 4:

[0642] The server uses the received audio feature data to perform sentiment analysis using a machine learning model. The analysis generates labels indicating emotions and identifies the user's emotional state.

[0643] Step 5:

[0644] The server generates response text that matches the original user statement based on the sentiment analysis results. Natural language processing techniques are used to select wording that reflects the user's emotions.

[0645] Step 6:

[0646] The server converts the generated response text into speech data using a speech synthesis engine. During this process, it adjusts intonation, pitch, and speed to express emotional nuances.

[0647] Step 7:

[0648] The server sends the synthesized audio data to the terminal. This transmission also involves data compression as needed.

[0649] Step 8:

[0650] The device plays the received audio data through its speaker. In this way, it provides the user with an emotionally resonant voice response.

[0651] (Example 1)

[0652] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0653] Traditional interaction technologies face the challenge of generating empathetic responses to user emotions in real time. Furthermore, if the generated responses do not adequately reflect the nuances of those emotions, communication with the user tends to become strained.

[0654] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0655] In this invention, the server includes means for acquiring the user's voice and converting it into an electrical signal, means for reading features from the electrical signal, and means for evaluating emotions using the read features. This enables the server to appropriately evaluate the user's emotions in real time, quickly generate empathetic responses based on those evaluations, and facilitate natural communication with the user.

[0656] "Voice" refers to the audio signals emitted by the user, which are collected as acoustic information.

[0657] An "electrical signal" is a format used for information processing, which is the conversion of sound into digital data.

[0658] A "feature" is a specific element of data extracted from an audio signal that contains information necessary for sentiment evaluation.

[0659] "Emotion" refers to the psychological state a user is in, and is identified through their vocal characteristics.

[0660] "Evaluation" is the process of determining the user's emotions based on features extracted from their voice.

[0661] "Response" refers to an audio or text message generated based on the user's sentiment evaluation.

[0662] "Synthesis" refers to the process of generating speech from text or data, and is a technology for providing voice responses to users.

[0663] "Output" refers to playing the generated audio through a device such as a speaker.

[0664] The emotion-empathizing AI agent system according to this invention uses multiple hardware and software components to provide natural interaction based on the user's emotions.

[0665] First, the terminal uses a smartphone or computer equipped with a microphone as an audio input device. The user's voice is picked up by the microphone as an analog signal and converted into an electrical signal on the spot. The electrical signal is sampled over a specific time interval through signal processing technology, and features are extracted as digital data. For this process, Mel-frequency cepstrum coefficients (MFCCs) are used as a method for frequency conversion.

[0666] Next, the server receives feature data sent from the terminal and uses a pre-trained learning algorithm to evaluate the user's emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used for emotion evaluation. As a result of the evaluation, emotion labels such as "joy," "sadness," and "anger" are assigned.

[0667] Based on the received emotion label, the server uses natural language processing techniques to generate an empathetic response. This response generation utilizes a generative AI model, which generates appropriate text in response to prompt text. For example, the prompt text "What should I say if the user's emotion is 'fatigue'?" is input to the model.

[0668] The generated text responses are converted into audio data using speech synthesis technology. Technologies such as WaveNet are used for speech synthesis to achieve natural and fluent expression. The audio data is then sent back to the terminal.

[0669] Ultimately, the device outputs synthesized speech data to the user through its speaker. This allows the user to naturally receive empathetic responses from the system, resulting in a more human-like interface.

[0670] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0671] Step 1:

[0672] The user emits sound through the microphone. This sound is input to the device and captured as an analog signal. Next, the device converts the analog signal into an electrical signal. The resulting electrical signal becomes the input for the audio data.

[0673] Step 2:

[0674] The terminal converts this electrical signal into digital audio data and samples it over a specific time interval. From the sampled digital data, it performs frequency conversion to extract Mel-frequency cepstrum coefficients (MFCCs). The MFCCs become the output feature data, representing important characteristics of the audio as numerical data.

[0675] Step 3:

[0676] The terminal sends the extracted MFCC data to the server as sample data. The server receives this data and uses it as input data for sentiment analysis. This transmitted feature data is the input for the next process.

[0677] Step 4:

[0678] The server uses the received feature data to run a deep learning model and evaluate emotions. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are used here. This outputs emotion labels. For example, emotions such as "joy," "anger," and "sadness" are analyzed.

[0679] Step 5:

[0680] The server generates empathetic response text using natural language processing techniques based on emotion labels. This process utilizes a generative AI model, which is input with prompts to output appropriate text. For example, a prompt might ask, "What should I say if the user's emotion is 'fatigue'?"

[0681] Step 6:

[0682] The server uses speech synthesis technology to convert the generated response text into speech. The generated speech data is synthesized using models such as WaveNet to produce output that reflects emotions.

[0683] Step 7:

[0684] The device receives synthesized speech data sent from the server and plays it through the speaker. Through this outputted speech, the user can receive emotionally resonant responses.

[0685] (Application Example 1)

[0686] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0687] Communicating with the elderly requires understanding their emotions appropriately and providing empathetic responses. However, conventional techniques are insufficient in adequately addressing the unique emotional states of the elderly. Therefore, there is a need for more effective communication methods to improve the quality of life for the elderly.

[0688] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0689] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, and means for generating responses that are empathetic to the emotions of elderly people based on the results of emotion analysis. This makes it possible to understand the unique emotional state of elderly people and to communicate appropriately with them.

[0690] "Sound" refers to the audio signal received from the user, which is the input data that the system uses to convert it into digital data and analyze it.

[0691] "Digital data" refers to data obtained by converting audio into a format that can be processed on a computer, and it serves as fundamental information used for feature extraction and sentiment analysis.

[0692] "Feature extraction" is the process of extracting specific information from audio data and deriving metrics necessary for data analysis. It is a crucial process that affects the accuracy of the analysis.

[0693] "Emotional analysis" is a process that analyzes the emotions contained in a user's utterances based on feature-extracted data and determines a specific emotional state.

[0694] "Response generation" is the process of creating appropriate text responses based on the results of sentiment analysis, enabling communication that is sensitive to the user's emotions.

[0695] "Speech synthesis" is the process of converting text-generated responses back into speech format, providing users with natural-sounding responses.

[0696] "Empathizing with the emotions of the elderly" means understanding the unique emotional states that elderly people exhibit, responding appropriately and reassuringly, and aiming to achieve comfortable communication.

[0697] To implement this invention, it is necessary to use a terminal that functions as a user interface and a server that performs voice data analysis. The terminal is equipped with a voice input device that collects the user's voice in real time and converts it into digital data.

[0698] The digital data converted by the terminal is sent to a server for feature extraction. The server uses software such as Python or LibROSA to extract Mel-frequency cepstrum coefficients. This allows the necessary features for sentiment analysis to be obtained from the audio data.

[0699] Sentiment analysis is performed on a server using a pre-trained deep learning model. This process utilizes machine learning frameworks such as TensorFlow to assign labels indicating specific emotions to audio data. Then, Hugging Face's Transformers are used to generate natural-sounding responses based on the analysis results.

[0700] The generated responses are converted into audio data using speech synthesis software such as Google Text-to-Speech. The converted audio data is sent to the terminal and played back to the user. This entire process enables the system to communicate in a way that is sensitive to the emotions of elderly people.

[0701] For example, if a user says, "My family isn't coming," the server can generate an empathetic response such as, "They are important family members, aren't they? Shall I think about what you'd like to talk about next time they visit?" An example of a prompt might be, "Generate a response that reflects the following emotion: Sadness - Appropriate comfort when the person you're talking to is sad."

[0702] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0703] Step 1:

[0704] The terminal collects the user's voice using an audio input device. The input is an analog audio signal, which is then converted into digital data. The converted digital data is then ready for transmission to the server.

[0705] Step 2:

[0706] The server processes the digital audio data received from the terminal and performs feature extraction. Specifically, it uses the LibROSA library to calculate Mel-frequency cepstrum coefficients. The input is digital audio data, and the output is feature data. This allows the characteristics of the audio to be quantified and used for sentiment analysis.

[0707] Step 3:

[0708] The server performs sentiment analysis based on feature data. A deep learning model (using TensorFlow) generates sentiment labels. The input is feature data, and the output is sentiment labels. These labels indicate the user's emotional state.

[0709] Step 4:

[0710] The server generates a response using natural language processing techniques based on the generated emotion labels. A text-based response is created using Hugging Face's Transformers. The input is the emotion label, and the output is the text response to be returned to the user.

[0711] Step 5:

[0712] The server converts text responses into speech data. It uses Google Text-to-Speech to synthesize speech from text. The input is the text response, and the output is synthesized speech data. This generates natural-sounding speech responses.

[0713] Step 6:

[0714] The terminal receives synthesized speech data from the server and plays it back to the user through the speaker. The input is synthesized speech data, and the output is the voice the user hears. This enables communication that is sensitive to the emotions of elderly people.

[0715] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0716] The emotion-empathizing AI agent system according to the present invention provides a new interactive system that combines an emotion engine to highly recognize the user's emotions.

[0717] First, the device uses its voice input function to collect the user's voice and converts the analog audio signal into digital data. In this process, the audio data is processed through noise filtering to make it clearer.

[0718] Next, the device extracts voice features from this digital data using Mel-frequency cepstrum coefficients (MFCCs), and in addition, collects user behavior data (e.g., gestures and behaviors), which it then sends to the emotion engine.

[0719] The server drives an emotion engine, integrating voice and behavioral data to analyze the user's emotions. A deep learning-based machine learning model is used for emotion recognition, which generates a variety of emotion labels. The emotion engine precisely identifies multiple emotional states, and the analysis results are stored on the server.

[0720] Based on the analysis results, the server utilizes natural language processing technology to generate responses that resonate with the user's emotions. The content and tone of the responses are dynamically adjusted based on the emotion recognition results. The generated response text is then converted into natural, emotionally expressive speech data by a speech synthesis engine.

[0721] Ultimately, the device plays synthesized speech to the user. This playback is done through the speaker, improving the user experience by providing responses appropriate to the situation the user is facing.

[0722] For example, if a user says "I'm sad today," the emotion engine immediately recognizes the emotion of "sadness." The server then generates a response of encouragement such as "You've had a tough day, please let me know if there's anything I can do to help," which is then played back via speech synthesis, providing the user with an empathetic interaction. In this way, the present invention provides a system that realizes a rich emotional experience for the user.

[0723] The following describes the processing flow.

[0724] Step 1:

[0725] The device collects the user's voice through a microphone. This voice is acquired as an analog signal and immediately converted into digital data. Furthermore, the voice data is processed using noise filtering to remove unwanted background noise and prepare it for voice analysis.

[0726] Step 2:

[0727] The device extracts features from digital audio data using Mel-frequency cepstrum coefficients (MFCCs). These features are numerical data that represent the characteristics of the voice and are used as basic data for speech recognition and sentiment analysis. The device also simultaneously collects user behavior data (e.g., gesture recognition via camera).

[0728] Step 3:

[0729] The terminal combines the extracted voice data and behavioral data and sends it to the server. The data is compressed during transmission to reduce network load.

[0730] Step 4:

[0731] The server analyzes the received data using an emotion engine. The emotion engine integrates voice and behavioral data using machine learning models to identify the user's emotional state. The identified emotions are categorized into specific labels, and the results are processed on the server.

[0732] Step 5:

[0733] The server applies a natural language processing algorithm based on the sentiment analysis results to generate the optimal response for the user. This response is designed with content and tone that resonates with the user's current emotions.

[0734] Step 6:

[0735] The server uses the generated response text to initiate the speech synthesis process. This process synthesizes speech that reflects emotional expression from the generated text, constructing the final audio file.

[0736] Step 7:

[0737] The server sends the synthesized audio to the terminal. The terminal also decodes the data as needed and prepares it to be delivered to the user in the optimal state.

[0738] Step 8:

[0739] The device plays the received synthesized speech through its speaker. This allows the user to hear responses in a voice and expression that reflects their own emotions, enabling them to enjoy a natural dialogue with the system.

[0740] (Example 2)

[0741] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0742] Conventional voice dialogue systems face the challenge of not being able to appropriately recognize user emotions and provide empathetic responses. Furthermore, it is necessary to perform more accurate emotion analysis by taking into account not only voice information but also behavioral information.

[0743] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0744] In this invention, the server includes means for acquiring user voice and converting it into digital information, means for extracting characteristics from the digital information, and means for acquiring user behavior and using it for emotion analysis. This enables integrated analysis of emotions from voice and behavior, and the generation of responses that are tailored to the user.

[0745] "User voice" refers to the human voice signal input into the system.

[0746] "Digital information" refers to data obtained by converting analog signals into a digital format.

[0747] "Characteristics" refer to distinctive elements or patterns contained in voice or behavioral data.

[0748] "Emotions" refer to information that indicates the user's psychological state.

[0749] A "response" is a message or action that a system returns to the user.

[0750] "Behavior" refers to information obtained from the user's gestures and actions.

[0751] The "frequency coefficient" is an index used to analyze the characteristics of audio data.

[0752] A "machine learning algorithm" is a technology that learns patterns from data and uses them to make predictions and classifications.

[0753] "Natural language processing technology" is a technology that enables computers to understand and generate human language.

[0754] In this invention, the user first inputs their voice into the system using a terminal equipped with an audio input device. The terminal is equipped with a high-performance microphone that can accurately capture the voice. The audio signal is converted from analog to digital, and noise is removed using digital signal processing technology. This results in clear audio data.

[0755] Next, the device uses signal processing software to extract speech characteristics by calculating Mel-frequency cepstrum coefficients (MFCCs) from the digitized speech data. It also captures user gestures and movements through cameras and motion sensors, collecting this as behavioral data. This consistent data, combined with speech, forms the basis for sentiment analysis.

[0756] The server uses a generative AI model powered by deep learning to evaluate integrated speech and behavioral data through sentiment analysis. This model is trained on a large amount of data and accurately identifies human emotions as multiple emotion labels. The identified emotions are stored on the server and updated as needed.

[0757] The server then uses natural language processing technology to generate a response that matches the user's emotional state. The generated response is adjusted in content and tone to be more natural and empathetic to the user's feelings. For example, if the user says, "I'm sad today," the system recognizes the emotion of "sadness" and generates a response such as, "You've had a tough day, please let me know if there's anything I can do to help."

[0758] The generated text is converted into expressive speech through a speech synthesis engine. The device then plays this synthesized speech through its speaker, delivering it directly to the user. This process allows the user to have an experience that feels like interacting with a human, rather than a robotic one.

[0759] Example of a prompt

[0760] "Today was very busy, and I'm a little tired. Could you tell me how to relax?"

[0761] "I'm feeling a little anxious about the new project. Could you give me some advice?"

[0762] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0763] Step 1:

[0764] The user inputs their voice into the device via the microphone. The device receives this voice as an analog signal and converts it to a digital signal using an analog-to-digital converter. This converted digital audio data is then subjected to a noise reduction filter to reduce background noise and obtain a clear result. This then becomes the input for the next step.

[0765] Step 2:

[0766] The device uses the cleared digital audio data to extract Mel-frequency cepstrum coefficients (MFCCs) using a signal processing algorithm. This analyzes the characteristics of the audio, which are then output as feature data. In addition, the camera and motion sensors capture the user's gestures and movements, which are stored as behavioral data. This data is then transmitted to the emotion analysis engine.

[0767] Step 3:

[0768] The server receives voice feature data and behavioral data sent from the terminal. Using a generative AI model, this data is integrated and sentiment analysis is performed. This analysis classifies the data into numerous sentiment labels, allowing for an understanding of the user's emotional state. This is crucial input for the next step.

[0769] Step 4:

[0770] The server uses natural language processing technology to generate user-specific responses based on the sentiment analysis results. The content and tone of the responses are adjusted according to the analyzed sentiment labels. The generated responses are output in text format.

[0771] Step 5:

[0772] The terminal receives generated text sent from the server and converts it into speech data using a speech synthesis engine. Speech synthesis produces speech with natural and emotional intonation, which becomes the final output.

[0773] Step 6:

[0774] The device plays synthesized speech to the user via its speaker. At this point, the user receives empathetic responses from the system in voice, completing the conversational experience.

[0775] (Application Example 2)

[0776] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0777] In elderly care settings, it is crucial to accurately understand changes in emotions and conditions, and to automatically provide empathetic dialogue and instructions accordingly. A lack of emotional understanding can lead to a decline in the satisfaction and quality of life of elderly individuals. A system is needed to address this situation while simultaneously reducing the burden on caregivers.

[0778] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0779] In this invention, the server includes means for collecting user voice and converting it into digital data, means for extracting features from the digital data, means for performing sentiment analysis on the extracted feature data and behavioral data, means for generating empathetic responses based on the analysis results, and means for synthesizing the generated responses as voice and making suggestions based on a specific context. This makes it possible to accurately grasp the emotional state of the person receiving care and provide empathetic and appropriate responses.

[0780] A "user" refers to a person who uses a system and interacts with it.

[0781] "Speech" refers to a waveform signal containing language emitted by the user.

[0782] "Digital data" is a data format that quantifies analog audio signals.

[0783] "Feature extraction" is the process of extracting meaningful patterns or information from digital data.

[0784] "Behavioral data" refers to information about a user's actions, gaze, gestures, and other behavioral data.

[0785] "Emotional analysis" is a process that evaluates a user's emotional state based on voice and behavioral data.

[0786] "Empathic response" refers to a response that aligns with and shows understanding of the user's emotions.

[0787] "Speech synthesis" is a technology that converts generated text into a speech format.

[0788] "Empathic and appropriate responses" refer to communication content that is relevant to the situation and emotions, generated based on the results of emotion analysis.

[0789] The system of the present invention is implemented as a care support tool using an emotion-empathetic AI agent. This system aims to improve the quality of care by understanding the emotions of the elderly and automatically generating empathetic and appropriate dialogue.

[0790] The server first collects the user's voice and converts the analog audio signal into digital data. The audio data is noise-filtered, and features are extracted using Mel-frequency cepstrum coefficients. In addition, user behavior data is collected. This data is sent to the emotion engine, where a deep learning-based machine learning model precisely analyzes the user's emotions.

[0791] Based on the analysis results, the server uses natural language processing technology to generate a user-centric response. This response is synthesized as natural speech by a speech synthesis engine and played back to the user through the speaker. Providing an appropriate response that is relevant to the user's current situation improves the user experience.

[0792] For example, if a user says, "I'm feeling a little down today," the emotion engine can identify the corresponding emotion, and the server can generate a response such as, "Is there anything I can do to help?" and deliver it via voice. Therefore, accuracy in emotion recognition and the ability to have natural conversations are considered important.

[0793] An example of a prompt to the generation AI model is, "Based on the user's voice, please understand their current emotions and come up with appropriate words of encouragement." This prompt enables the system to generate and provide emotionally resonant and appropriate dialogue.

[0794] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0795] Step 1:

[0796] The device collects the user's voice in real time through the microphone. The collected analog audio signal is converted into digital data. This digital data is then filtered to remove unwanted sounds and used as input data.

[0797] Step 2:

[0798] The device extracts speech features from noise-filtered digital data using Mel-frequency cepstrum coefficients (MFCCs). The output is feature data that shows the frequency pattern of the speech.

[0799] Step 3:

[0800] The device collects user behavior data as additional input. This behavior data includes gestures and actions, and is also recorded in digital format.

[0801] Step 4:

[0802] The device sends voice feature data and behavioral data to the server. The server inputs this data into an emotion engine and analyzes the user's emotions using a generative AI model. The output of this step is the identified emotion label.

[0803] Step 5:

[0804] The server uses natural language processing technology to generate empathetic responses tailored to the user based on the analyzed emotion labels. The responses are output in text format, and their content is relevant to the user's situation and emotions.

[0805] Step 6:

[0806] The server inputs the generated text response into a speech synthesis engine, which converts it into natural and expressive speech data. The output is the synthesized speech file.

[0807] Step 7:

[0808] The device plays synthesized speech to the user through its speaker. This playback process allows the user to receive empathetic messages at appropriate times.

[0809] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0810] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0811] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[0812] Furthermore, the emotion identification model 59, acting as an emotion engine, may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0813] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[0814] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[0815] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[0816] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[0817] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[0818] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[0819] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[0820] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[0821] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0822] 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.

[0823] Furthermore, it is not necessary to store the entirety of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[0824] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[0825] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[0826] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[0827] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[0828] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[0829] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0830] The following is further disclosed regarding the embodiments described above.

[0831] (Claim 1)

[0832] A means of collecting user voice and converting it into digital data,

[0833] A means for extracting features from the digital data,

[0834] A method for performing sentiment analysis on feature-extracted data,

[0835] A means for generating an optimal response based on the analysis results,

[0836] A means for synthesizing the generated response as speech,

[0837] A means of playing back synthesized speech to the user,

[0838] A system that includes this.

[0839] (Claim 2)

[0840] The system according to claim 1, wherein the speech analysis means utilizes Mel-frequency cepstrum coefficients for the purpose of feature extraction.

[0841] (Claim 3)

[0842] The system according to claim 1, wherein a machine learning model is used in the response generation means based on the analysis results.

[0843] "Example 1"

[0844] (Claim 1)

[0845] A means of acquiring user voices and converting them into electrical signals,

[0846] A means for reading characteristics from the electrical signal,

[0847] A method for evaluating emotions using the characteristics that have been read,

[0848] A means of creating empathetic responses in response to evaluation results,

[0849] A means of synthesizing the created response as sound,

[0850] A means of outputting the synthesized sound to the user,

[0851] A system that includes this.

[0852] (Claim 2)

[0853] The system according to claim 1, wherein the audio analysis means utilizes frequency conversion for the purpose of reading the aforementioned features.

[0854] (Claim 3)

[0855] The system according to claim 1, wherein a learning algorithm is used in the response generation means based on the evaluation results.

[0856] "Application Example 1"

[0857] (Claim 1)

[0858] A means of collecting user voice and converting it into digital data,

[0859] A means for extracting features from the digital data,

[0860] A method for performing sentiment analysis on feature-extracted data,

[0861] A means for generating an optimal response based on the analysis results,

[0862] A means for synthesizing the generated response as speech,

[0863] A means of playing back synthesized speech to the user,

[0864] A means of generating responses that are empathetic to the emotions of elderly people, based on the results of emotion analysis,

[0865] A system that includes this.

[0866] (Claim 2)

[0867] The system according to claim 1, wherein the speech analysis means utilizes Mel-frequency cepstrum coefficients for the purpose of feature extraction.

[0868] (Claim 3)

[0869] The system according to claim 1, wherein a machine learning model is used in a response generation means based on analysis results to generate a response that corresponds to the specific emotional state of elderly people.

[0870] "Example 2 of combining an emotion engine"

[0871] (Claim 1)

[0872] A means of acquiring user voice and converting it into digital information,

[0873] A means for extracting characteristics from the digital information,

[0874] A method for analyzing emotions from extracted information,

[0875] A means for generating the optimal response based on the results of emotion analysis,

[0876] A means for synthesizing the generated response as speech,

[0877] A means of presenting synthesized speech to the user,

[0878] A means of acquiring user behavior and using it for sentiment analysis,

[0879] A means for comprehensively analyzing emotions from voice and behavior,

[0880] A means for dynamically adjusting the content and tone of responses using natural language processing technology,

[0881] A system that includes this.

[0882] (Claim 2)

[0883] The system according to claim 1, wherein the audio analysis means uses frequency coefficients for the purpose of extracting the characteristics.

[0884] (Claim 3)

[0885] The system according to claim 1, wherein a machine learning algorithm is applied to the response generation means based on the analysis results.

[0886] "Application example 2 when combining with an emotional engine"

[0887] (Claim 1)

[0888] A means of collecting user voice and converting it into digital data,

[0889] A means for extracting features from the digital data,

[0890] A method for performing sentiment analysis on feature-extracted data and behavioral data,

[0891] A means for generating empathetic responses based on analysis results,

[0892] A means for synthesizing the generated response as speech,

[0893] A means of playing synthesized speech to the user and making suggestions based on a specific context,

[0894] A system that includes this.

[0895] (Claim 2)

[0896] The system according to claim 1, wherein the speech analysis means utilizes Mel-frequency cepstrum coefficients for the purpose of feature extraction.

[0897] (Claim 3)

[0898] The system according to claim 1, wherein a machine learning model is used in the response generation means based on the analysis results. [Explanation of symbols]

[0899] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting user voice and converting it into digital data, A means for extracting features from the digital data, A method for performing sentiment analysis on feature-extracted data, A means for generating an optimal response based on the analysis results, A means for synthesizing the generated response as speech, A means of playing back synthesized speech to the user, A means of generating responses that are empathetic to the emotions of elderly people, based on the results of emotion analysis, A system that includes this.

2. The system according to claim 1, wherein the speech analysis means utilizes Mel-frequency cepstrum coefficients for the purpose of feature extraction.

3. The system according to claim 1, wherein a machine learning model is used in the response generation means based on the analysis results to generate a response that corresponds to the specific emotional state of elderly people.

Images