System

Abstract

A system is provided.SOLUTION: A system comprising: means for inputting voice data; means for analyzing the input voice data in real time and extracting a feature amount of a speaker; means for specifying a habit of the speaker based on the extracted feature amount and removing the habit; means for regenerating the voice data from which the habit is removed as a natural utterance; and means for reproducing the regenerated voice data to a user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] Habits of rapid speech, filler words (meaningless words), and stammering that occur when people speak can prevent the speaker from accurately conveying their intentions to the listener. This is a particularly serious problem for people who stutter or who feel anxious about their speaking style, leading to missed communication opportunities. These habits can also make smooth communication difficult in business and personal situations. There is a need to provide a system that can solve these issues and accurately convey the speaker's original intentions. [Means for solving the problem]

[0005] The present invention provides a system including: means for inputting speech data; means for analyzing the input speech data in real time and extracting features of the speaker; means for identifying the speaker's habits based on the extracted features and removing the habits; means for regenerating the speech data from which the habits have been removed as natural speech; and means for playing back the regenerated speech data to the user. This system analyzes the speech data in real time and removes features such as rapid speech, stammers, and fillers, thereby playing back natural speech that accurately conveys the speaker's intentions. This allows even people who stutter or are anxious about their speaking style to communicate with peace of mind. Furthermore, by performing noise reduction and volume adjustment as preprocessing on the speech data, even clearer speech can be provided.

[0006] "Voice data input means" refers to a device or software for acquiring voice from a user and inputting it into the system.

[0007] "Means for analyzing in real time" refers to algorithms and hardware that instantly analyze captured audio and process it without delay.

[0008] "Means for extracting features" refers to technology for extracting and analyzing speech habits and characteristics as data from voice data.

[0009] "Means for identifying speaker habits" refers to technology that identifies habits contained in a speaker's speech, such as rapid speech, stammering, and fillers, based on extracted features.

[0010] "Habit removal methods" refers to generative AI techniques that modify and remove identified speaker habits to produce natural, fluent speech.

[0011] "Means for regenerating natural speech" refers to a technology that regenerates speech data after removing habits as natural, easy-to-listen-to speech for the user.

[0012] "Means for playing to a user" refers to a device or software for playing the regenerated audio data to a user.

[0013] "Noise reduction" refers to technology that reduces background noise from audio data to provide clearer audio.

[0014] "Volume adjustment" refers to a technology that adjusts the volume of audio data to an appropriate level to provide audio that is easy to hear. [Brief explanation of the drawings]

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

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

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

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

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

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

[0021] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0022] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0023] [First embodiment]

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

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

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

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

[0028] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

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

[0030] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

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

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

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

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

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

[0036] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used with devices such as telephones, web conferences, and earphones with microphones. Specific embodiments of the present invention are described below.

[0037] System Configuration

[0038] The system consists of the following main components:

[0039] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[0040] Server: A processing device that receives voice data, analyzes it in real time, removes habits, and reproduces natural speech.

[0041] User: The entity that inputs speech and receives modified, natural-speech audio.

[0042] Program processing

[0043] The operation of the system will be specifically described below.

[0044] Input and transmission of voice data

[0045] 1. The user inputs audio using earphones with a microphone or a PC microphone. This audio is normal conversation.

[0046] 2. The device captures the audio data in real time and prepares to send it to the server.

[0047] 3. The device sends the captured audio data to the server.

[0048] Analyzing voice data and identifying habits

[0049] 4. The server receives the voice data sent from the device.

[0050] 5. The server performs preprocessing on the received audio data, including noise reduction and volume adjustment.

[0051] 6. The server extracts features from the audio data, including features such as rapid speech, stammers, and fillers.

[0052] 7. The server identifies the speaker's habits based on the extracted features.

[0053] Removes quirks and reproduces natural speech

[0054] 8. The server applies generative AI models to remove identified habits, e.g., remove fillers and convert fast-paced parts to an appropriate speed.

[0055] 9. The server recreates natural-sounding speech based on the data after the habits have been removed.

[0056] Sending and playing back modified audio

[0057] 10. The server sends the regenerated audio data to the device.

[0058] 11. The device plays the modified audio data to the user, allowing the user to hear natural, fluent speech.

[0059] Specific examples

[0060] A specific example of use is shown below.

[0061] Call Scenarios

[0062] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0063] The device captures this audio and transmits it to the server in real time.

[0064] The server analyzes the audio data and identifies fillers such as "um" and "right."

[0065] The server removes the filler and converts it into natural speech, such as "I'd like to ask for a moment of your time."

[0066] The server sends the regenerated audio to the terminal, which plays it back.

[0067] The user can continue the conversation as if they were speaking fluently while listening to the other party's responses.

[0068] This invention enables users who are unsure about their speaking style or who stutter to communicate naturally and smoothly. This system will be an extremely useful tool in both business and personal settings.

[0069] The processing flow will be explained below.

[0070] Specific processing steps of the program

[0071] Step 1:

[0072] The user inputs audio using earphones with a microphone or the microphone on the PC.

[0073] The terminal captures the audio data in real time and stores the data in a buffer.

[0074] Step 2:

[0075] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0076] The device sends the audio data to the server in real time whenever possible.

[0077] Step 3:

[0078] The server receives the voice data transmitted from the terminal.

[0079] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[0080] Step 4:

[0081] The server adjusts the volume to set the optimal listening level.

[0082] The server completes preprocessing in order to analyze the audio data in real time.

[0083] Step 5:

[0084] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0085] The server analyzes the extracted features and identifies speaking habits.

[0086] Step 6:

[0087] The server generates a dataset to correct the identified habits.

[0088] The server applies a generative AI model to make corrections such as adjusting fast-talking parts and removing filler.

[0089] Step 7:

[0090] The server recreates natural speech based on the modified voice data.

[0091] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0092] Step 8:

[0093] The server transmits the regenerated voice data to the terminal in real time.

[0094] The terminal receives the modified natural speech audio and plays it back to the user.

[0095] Specific examples

[0096] Call Scenarios

[0097] Step 1:

[0098] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0099] The device captures this audio in real time and stores it in a buffer.

[0100] Step 2:

[0101] The device sends the captured audio data to the server.

[0102] Step 3:

[0103] The server receives the audio data and performs noise reduction.

[0104] Step 4:

[0105] The server adjusts the volume.

[0106] Step 5:

[0107] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0108] The server identifies the speech habits based on the identified feature amount.

[0109] Step 6:

[0110] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[0111] Step 7:

[0112] The server reproduces the natural utterance "Please give me a moment."

[0113] Step 8:

[0114] The server sends the regenerated audio to the terminal, which plays it back to the user.

[0115] Example 1

[0116] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0117] Conventional speech data analysis systems have had difficulty accurately removing speaker habits and regenerating natural speech. In particular, there is a need to provide natural conversations by removing characteristic habits such as rapid speech and fillers in real time. To solve this problem, an efficient and accurate method for analyzing and regenerating speech data is required.

[0118] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0119] In this invention, the server includes means for receiving speech data and performing preprocessing such as noise reduction and volume adjustment, means for extracting features from the speech data and identifying the speaker's habits, and means for applying a generative model to remove the identified habits and regenerate natural speech. This makes it possible to generate natural speech in real time while removing the speaker's unnatural habits.

[0120] "Voice data" refers to voice information stored in digital format that a user makes through an input device such as a microphone.

[0121] "Real-time" refers to data and events being processed as they occur, with immediate results available without delay.

[0122] "Noise reduction" is a technology for improving the clarity of audio by removing unnecessary background sounds and noise from audio data.

[0123] "Volume adjustment" is a process for equalizing the volume level of recorded audio data to ensure that it is easy to listen to when played back.

[0124] "Features" refer to specific patterns or attributes extracted from speech data, and are information used to identify a speaker's habits and characteristics.

[0125] A "generative model" is an artificial intelligence model used to create new data from existing data, and in this case is specifically used to regenerate audio data.

[0126] "Fillers" refer to unnecessary parts of speech, such as "ums" and "hmms," that speakers unconsciously use to connect words.

[0127] "Habits" refer to a speaker's particular speech patterns or habitual speaking characteristics that need to be identified and eliminated.

[0128] "Regeneration" refers to the process of reconstructing processed data in its original or new form, which in this case means producing natural-sounding speech.

[0129] This system analyzes input speech data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used on devices such as telephones, web conferences, and earphones with microphones.

[0130] System Configuration

[0131] The system consists of the following main components:

[0132] Terminal: A device that allows a user to input voice and transmit the voice data to a server. Examples include smartphones, PCs, and tablets.

[0133] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and regenerates natural speech. The server is equipped with a high-performance processor and sufficient memory.

[0134] User: The entity that inputs speech and receives modified, natural-speech audio.

[0135] Hardware and Software

[0136] The following hardware and software are used to implement this system.

[0137] Hardware:

[0138] Earphones with a microphone or a microphone built into your PC: Captures your voice.

[0139] Devices such as smartphones, PCs, and tablets: Captures audio data in real time and sends it to a server.

[0140] Server equipped with a high-performance processor: Analyzes and processes voice data.

[0141] software:

[0142] Audio processing library "SoX": Performs preprocessing for noise reduction and volume adjustment.

[0143] Machine learning library "TensorFlow": Extracts features from audio data and identifies habits.

[0144] Generative AI model "GPT-3": Removes fillers and reproduces natural speech.

[0145] "Vosk" speech generation library: converts text data into acoustic data.

[0146] Specific examples

[0147] Below are some specific examples of how the system can be used.

[0148] Call Scenarios

[0149] 1. During a call, the user says, "Um, I'd like to have a moment of your time."

[0150] 2. The device captures this audio and sends it to the server in real time.

[0151] 3. The server analyzes the audio data and identifies fillers such as "uh" and "right."

[0152] 4. The server removes the filler and converts the utterance into natural speech: "Please give me a moment of your time."

[0153] 5. The server sends the regenerated audio to the device, which plays it back.

[0154] 6. Users can continue the conversation as if they were speaking fluently while listening to the responses of the other party.

[0155] Prompt Sentence Examples

[0156] Below are some examples of prompts that can be given to the generative AI model.

[0157] Prompt Sentence Examples

[0158] Input: "Um, I'd like to, well, take a moment."

[0159] Instructions to the generative AI model: "Remove fillers and convert to natural-sounding speech."

[0160] In this way, a system is provided that allows the speaker to have a smooth conversation.

[0161] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0162] Step 1:

[0163] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is a normal conversation, such as "Um, I'd like to have a moment of your time." The input voice data is captured as an analog signal and converted to a digital signal within the device.

[0164] Step 2:

[0165] The device stores the captured audio data in memory in real time and prepares it for transmission to the server. During this process, the audio data is divided into packets at short time intervals (for example, every second). The data is then sent to the server using a secure communication protocol such as HTTPS. The input is analog audio data, and the output is packetized digital audio data.

[0166] Step 3:

[0167] The server receives the audio data sent from the terminal. The received packets are reassembled to form a continuous audio data stream. The input is packetized digital audio data, and the output is a reassembled continuous audio data stream.

[0168] Step 4:

[0169] The server performs pre-processing for noise reduction and volume adjustment on the received audio data. This process uses the "SoX" library to remove background noise and equalize the volume. Specifically, it applies noise filtering and audio normalization algorithms. The input is a reconstructed continuous audio data stream, and the output is pre-processed, clean audio data.

[0170] Step 5:

[0171] The server extracts features from the preprocessed audio data. In this step, the machine learning library "TensorFlow" is used to calculate features such as Mel-Frequency Cepstrum Coefficients (MFCCs) from the audio data. The features represent specific attributes of the audio data and are used as input data to identify the speaker's habits. The input is preprocessed, clean audio data, and the output is the extracted audio features.

[0172] Step 6:

[0173] The server identifies the speaker's habits based on the extracted features. This process uses an algorithm trained on previous datasets to detect speech habits such as quick speech, stammers, and fillers. The input is the speech features, and the output is the identified speaker's habits.

[0174] Step 7:

[0175] The server applies a generative model to remove the identified habits. Specifically, it uses the GPT-3 model and provides a prompt, such as "Please remove fillers and convert to natural speech." The generative model removes unnecessary fillers and habits and converts the speech to natural speech. The input is speaker habit information and audio data, and the output is natural speech data with the habits removed.

[0176] Step 8:

[0177] The server recreates natural speech data with the quirks removed. In this step, the "Vosk" library is used to convert text data into acoustic data and generate natural speech. The input is the identified audio-text data, and the output is the recreated continuous audio data.

[0178] Step 9:

[0179] The server sends the regenerated audio data to the device, where it is repacketized and transmitted securely via a protocol such as HTTPS. The input is the regenerated continuous audio data, and the output is packetized digital audio data.

[0180] Step 10:

[0181] The terminal receives the modified audio data and plays it back to the user. Specifically, it reconstructs the received packetized data and plays it back to the user through a speaker or earphone. The input is packetized digital audio data, and the output is played-back audio that the user can hear.

[0182] (Application example 1)

[0183] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0184] In a typical store, if a store clerk has a habit of speaking in a certain way when interacting with customers, it can cause discomfort to the customer. In particular, speaking too quickly, stammering, and using too many fillers can be obstacles to smooth communication with customers. Unless this issue is resolved, there is a risk that the quality of store service will decline and customer satisfaction will decrease. Therefore, the present invention aims to provide a system that automatically removes store clerks' habitual speech patterns and converts them into natural, fluent speech, thereby facilitating communication between store clerks and customers and improving customer satisfaction.

[0185] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0186] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, and means for analyzing and correcting the voice of the store clerk to smoothly serve customers in the store. This enables the store clerk to speak naturally and fluently without worrying about their speaking habits.

[0187] "Audio data" refers to data in which audio information is recorded in digital format.

[0188] "Analyze in real time" means processing data as it is entered.

[0189] "Speaker features" are data that represent specific patterns and attributes contained in the speaker's voice.

[0190] "De-embedding" means removing or modifying unwanted features in a speaker's speech (e.g., tongue twisters, fillers, etc.).

[0191] "Natural speech" refers to fluent and smooth speech that does not sound strange to the listener.

[0192] "Playback" means outputting the processed audio data as sound through speakers or headphones.

[0193] A "system" is a structure that includes multiple components that work in conjunction with each other.

[0194] A "store" is a physical location for selling goods and services.

[0195] "Customer service" refers to the service and interaction activities that store staff engage in with customers.

[0196] "Modification" means to change the original state and adjust it to a desired form.

[0197] The present invention is a system that analyzes voice data of store clerks in real time, automatically removes speaking habits, and reproduces natural speech in order to facilitate smooth customer service in stores. This system captures the voice of the store clerk using a device such as a smartphone or earphones with a microphone, and processes the voice data on a server. Specific embodiments of the present invention are described below.

[0198] System Configuration

[0199] The system consists of the following main components:

[0200] Terminal: A device that allows the store clerk to input voice and transmit the voice data to the server. Specifically, a smartphone or earphones with a microphone are used.

[0201] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and reproduces natural speech.

[0202] Store Clerk: The subject who inputs speech and receives modified natural speech audio.

[0203] Program Processing Overview

[0204] The program of this system executes the following processes.

[0205] 1. The terminal captures the clerk's voice in real time and sends it to the server.

[0206] 2. The server receives the audio data and performs preprocessing such as noise reduction and volume adjustment.

[0207] 3. The server extracts features from the speech data, including features such as rapid speech, stammers, and fillers.

[0208] 4. The server identifies the speaker's habits based on the extracted features.

[0209] 5. The server applies generative AI models to remove quirks, e.g., remove fillers and convert fast-paced parts to the appropriate speed.

[0210] 6. The server recreates natural-sounding speech based on the data after the habits have been removed.

[0211] 7. The server sends the regenerated audio data to the device.

[0212] 8. The terminal plays the corrected voice data to the store clerk.

[0213] Hardware used

[0214] Smartphone: A device for voice input and playback.

[0215] Earphones with microphone: A device for high-quality audio capture.

[0216] Software used

[0217] Python: The base programming language.

[0218] SpeechRecognition: A library for speech capture and recognition.

[0219] requests: An HTTP client library for sending audio data to a server.

[0220] Specific examples

[0221] For example, if a store clerk says, "Um, well, I'd like to have a moment of your time," the device captures this speech and sends it to the server in real time. The server analyzes the speech data, identifies fillers such as "um" and "I see," and removes them to convert it into natural speech, "I'd like to have a moment of your time." The server then sends the regenerated speech data to the device, which plays it back to the store clerk. This allows the store clerk to interact with the customer as if they were speaking fluently.

[0222] Prompt Sentence Examples

[0223] For the generative AI model that "removes fillers" and "converts to natural speech," use prompts like these:

[0224] It analyzes the voice data, identifies and removes unnecessary fillers (such as "um" or "hmm"), converts the speech into natural-sounding speech, and converts fast-paced speech to an appropriate speed, outputting smooth speech.

[0225] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0226] Step 1:

[0227] The user inputs voice using earphones with a microphone or the microphone on a smartphone. At this time, the user conducts normal conversation and the voice data is captured. The input voice is stored as raw data on the device.

[0228] Step 2:

[0229] The device transmits the captured audio data to the server in real time. Specifically, the device converts the audio data into a digital format and sends the data to the server using an HTTP request, with the audio data remaining raw.

[0230] Step 3:

[0231] The server receives the voice data sent from the device, converts the received data into an analyzable format, and uses it in the next processing step.

[0232] Step 4:

[0233] The server preprocesses the audio data. Specifically, it performs noise reduction and volume adjustment. This processing improves the quality of the audio data, making it easier to perform subsequent analysis. The input is raw audio data, and the output is audio data with noise reduced and volume adjusted.

[0234] Step 5:

[0235] The server extracts speaker features from the audio data. For example, it uses a speech recognition library to identify features such as rapid speech, stammers, and fillers from the audio signal. At this stage, features are generated as numerical data or in the form of tags.

[0236] Step 6:

[0237] The server identifies the speaker's habits based on the extracted features. For example, if there are many fillers, it will tag them as a "filler habit," and if they speak quickly, it will identify them as a "fast speaking habit." In this step, the habit data is output as the result of feature analysis.

[0238] Step 7:

[0239] The server applies a generative AI model to remove the identified habits. Specifically, it modifies the voice data using prompts for the generative AI model ("Analyze the voice data, identify and remove unnecessary fillers (such as 'um' or 'hmm'), and convert it into natural speech. Convert fast-paced parts to an appropriate speed and output smooth speech."). The input is the voice data with identified habits, and the output is the modified natural speech data.

[0240] Step 8:

[0241] The server sends the voice data regenerated into natural speech to the device. The modified voice data is sent back to the device using an HTTP request. The input is the natural speech data, and the output is the completion status of transmission to the device.

[0242] Step 9:

[0243] The terminal plays the corrected voice data to the user. The clerk can hear fluent and natural speech through a playback device (earphones or smartphone speakers). The input is the corrected voice data received from the server, and the output is the voice that the user can hear.

[0244] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0245] The present invention relates to a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. The system further includes an emotion engine that recognizes the user's emotions and adjusts the method of modifying the voice data based on the emotions.

[0246] System Configuration

[0247] The system consists of the following main components:

[0248] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[0249] Server: A processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech.

[0250] User: The entity that inputs speech and receives modified, natural-speech audio.

[0251] Emotion engine: A module that recognizes the user's emotions in real time and feeds the analysis results back to the system.

[0252] Program processing

[0253] The specific processing of the system will be explained below.

[0254] Input and transmission of voice data

[0255] The user inputs voice using earphones with a microphone or a PC microphone. This voice is normal conversation.

[0256] The terminal captures the audio data in real time and stores the data in a buffer.

[0257] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0258] The device sends the audio data to the server in real time whenever possible.

[0259] Voice data analysis and emotion recognition

[0260] The server receives the voice data transmitted from the terminal.

[0261] The server performs pre-processing of the received audio data, including noise reduction and volume adjustment.

[0262] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0263] The server analyzes the extracted features and identifies speaking habits.

[0264] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[0265] Habit removal and emotional adjustment

[0266] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[0267] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0268] Reproduction and playback of spontaneous speech

[0269] The server recreates natural speech based on the modified voice data.

[0270] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0271] The server transmits the regenerated voice data to the terminal in real time.

[0272] The terminal receives the modified natural speech audio and plays it back to the user.

[0273] Specific examples

[0274] Call Scenarios

[0275] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0276] The device captures this audio in real time and stores it in a buffer.

[0277] The device sends the captured audio data to the server.

[0278] The server receives the audio data and performs noise reduction and volume adjustment.

[0279] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0280] The server identifies the speech habits based on the identified feature amount.

[0281] The emotion engine recognizes that the user is feeling a little nervous and sends that information to the server.

[0282] The server will remove identified fillers such as "uh" and "right," convert fast-paced parts to an appropriate speed, and adjust the volume slightly to provide a sense of security.

[0283] The server reproduces the natural utterance "Please give me a moment."

[0284] The server sends the regenerated audio to the terminal, which plays it back to the user.

[0285] This invention enables natural and smooth communication even for users who are unsure about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses are realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[0286] The processing flow will be explained below.

[0287] Specific processing steps of the program

[0288] Step 1:

[0289] The user inputs audio using earphones with a microphone or the microphone on the PC.

[0290] The terminal captures the audio data in real time and stores the data in a buffer.

[0291] Step 2:

[0292] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0293] The device sends the audio data to the server in real time whenever possible.

[0294] Step 3:

[0295] The server receives the voice data transmitted from the terminal.

[0296] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[0297] The server adjusts the volume to set the optimal listening level.

[0298] Step 4:

[0299] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0300] The server analyzes the extracted features and identifies speaking habits.

[0301] Step 5:

[0302] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[0303] The emotion engine feeds back the recognized emotion information to the server.

[0304] Step 6:

[0305] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[0306] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0307] Step 7:

[0308] The server reproduces natural-spoken speech with the quirks removed and the emotions reflected.

[0309] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0310] Step 8:

[0311] The server transmits the regenerated voice data to the terminal in real time.

[0312] The terminal receives the modified natural speech audio and plays it back to the user.

[0313] Specific examples

[0314] Call Scenarios

[0315] Step 1:

[0316] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0317] The device captures this audio in real time and stores it in a buffer.

[0318] Step 2:

[0319] The device sends the captured audio data to the server.

[0320] Step 3:

[0321] The server receives the audio data and performs noise reduction and volume adjustment.

[0322] Step 4:

[0323] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0324] The server identifies the speech habits based on the identified feature amount.

[0325] Step 5:

[0326] The emotion engine recognizes that the user is feeling a little nervous.

[0327] The emotion engine sends the recognized emotion information to the server.

[0328] Step 6:

[0329] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[0330] Based on feedback from the emotion engine, the server makes adjustments such as slowing down the speaking rate and lowering the volume.

[0331] Step 7:

[0332] The server reproduces the natural utterance "Please give me a moment."

[0333] The server converts the reproduced audio data into a clear, audible format.

[0334] Step 8:

[0335] The server transmits the regenerated voice data to the terminal.

[0336] The terminal plays the modified natural speech to the user.

[0337] Example 2

[0338] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0339] Conventional speech recognition systems have had the problem of being unable to completely remove unnatural speech caused by the speaker's unique habits and emotions. Furthermore, they were unable to recognize the speaker's emotions and adjust their speech accordingly, making it difficult to achieve more natural and personalized speech communication. This has made it difficult for users who are unsure about their speaking style or who stutter to communicate smoothly.

[0340] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0341] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time, means for extracting features from the analyzed voice data, means for identifying the speaker's habits based on the extracted features and removing the habits, means for recognizing the user's emotions in real time and sending the information to the analysis means, means for regenerating the voice data from which the habits have been removed and which has been adjusted based on the emotion information as natural speech, and means for playing back the regenerated voice data to the user. This makes it possible to adjust and optimize the voice data based on the speaker's habits and emotions, thereby achieving natural and smooth communication.

[0342] "Means for inputting voice data" refers to equipment or a method by which a user inputs voice data into the system using a microphone or other input device.

[0343] "Means for real-time analysis" refers to algorithms or software that process and analyze input voice data almost instantaneously.

[0344] "Feature extraction" refers to a process for extracting specific patterns or attributes (e.g., speaking rate, fillers, stammers, etc.) from speech data.

[0345] "Means for identifying and removing speaker habits" refers to technology that detects a speaker's specific habits (e.g., speaking quickly, using fillers, etc.) based on extracted features and corrects or removes them.

[0346] The "means for recognizing the user's emotions in real time and transmitting that information to the analysis means" is a module for identifying the user's emotional state (e.g., joy, anger, sadness, surprise, etc.) from voice data in real time and reflecting that information in voice analysis.

[0347] A "regenerating means" is an algorithm or device that uses the modified speech data to reconstruct natural speech.

[0348] The "playback means" refers to an audio output device such as a speaker or earphones that allows the user to hear the regenerated audio data.

[0349] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, regenerates natural speech, and further recognizes the user's emotions and corrects the voice data based on those emotions. An embodiment of the present invention consists of the following main components and processing flow.

[0350] System Configuration

[0351] 1. Terminal

[0352] A device that allows a user to input voice and transmit the voice data to a server. Specific examples include earphones with a microphone and a microphone connected to a PC.

[0353] 2. Server

[0354] It is a processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech. The server processes the voice data using an advanced generative AI model.

[0355] 3. Emotion Engine

[0356] This module recognizes the user's emotions in real time and feeds the analysis results back to the system. The emotion engine extracts emotions such as joy, anger, sadness, and surprise from voice data.

[0357] 4. Users

[0358] A subject that inputs speech and receives modified, natural-speech audio.

[0359] Program processing

[0360] Input and transmission of voice data

[0361] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is captured in real time by the device and stored in a buffer. The device then prepares the voice data for transmission and starts sending it to the server. Transmission is performed as quickly as possible in real time.

[0362] Receiving and preprocessing audio data

[0363] The server receives the audio data sent from the device, applies a noise reduction filter to the received audio data to remove background noise, and performs volume equalization to adjust and maintain a constant volume level.

[0364] Voice data analysis and emotion recognition

[0365] The server extracts features from the preprocessed voice data. These features include speaking rate, fillers (e.g., "um," "ah," etc.), and stammers. The server analyzes the features of the voice data and identifies speaking habits. At the same time, the emotion engine recognizes the user's emotions from the voice data in real time and sends that information to the server.

[0366] Habit removal and emotional adjustment

[0367] The server generates a dataset for removing habits based on the identified habits and emotional information. Specifically, it uses a generative AI model to modify the speech data, removing identified fillers and adjusting fast-paced parts to an appropriate speed. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0368] Reproduction and playback of spontaneous speech

[0369] The server uses the modified voice data to recreate natural-sounding speech. The recreated voice data is passed through a sound quality optimization engine, which converts it into a clear, easy-to-listen format. Finally, the server transmits the optimized voice data to the device in real time, and the device plays the received voice back to the user.

[0370] Specific examples

[0371] Call Scenarios

[0372] During a call, a user utters, "Um, well, I'd like to have a moment of your time." The device captures this audio in real time and stores it in a buffer. The device then sends the stored audio data to the server. The server receives the audio data and performs noise reduction and volume adjustment. Features such as speech rate, fillers, and stammers are extracted from the audio data and speech habits are identified based on these. At the same time, the emotion engine recognizes whether the user is feeling nervous and sends this information to the server. The server then removes the identified fillers, adjusts the fast-talking parts to an appropriate speed, and adjusts the volume to provide a sense of security. The regenerated, natural utterance, "I'd like to have a moment of your time," is then sent to the device, which plays it back to the user.

[0373] This system enables natural and smooth communication even for users who are anxious about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses can be realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[0374] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0375] Step 1: Input audio data

[0376] The user inputs voice using earphones with a microphone or the microphone on the PC. At this stage, the user can have a normal conversation.

[0377] Input: User speaking

[0378] Output: Analog audio signal

[0379] Step 2: Capturing and buffering audio data

[0380] The terminal captures the user's input voice in real time and converts it into a digital voice signal.

[0381] The device temporarily stores captured digital audio in a buffer.

[0382] Input: Analog audio signal

[0383] Output: Digital audio data

[0384] Step 3: Prepare to send audio data

[0385] The terminal reads the voice data stored in the buffer at regular intervals and prepares it for transmission, during which preprocessing such as data compression is performed.

[0386] Input: Buffered digital audio data

[0387] Output: Audio data format that can be sent to the server

[0388] Step 4: Sending audio data

[0389] The data is converted into a voice data format that can be transmitted by the terminal and sent to the server. Transmission is performed as quickly as possible in real time.

[0390] Input: Audio data format that can be sent to the server

[0391] Output: Audio data sent to the server

[0392] Step 5: Receiving audio data

[0393] The server receives the voice data transmitted from the terminal in real time.

[0394] Input: Audio data sent from the device

[0395] Output: Received audio data stored on the server

[0396] Step 6: Preprocessing the audio data

[0397] The server applies a noise reduction filter to remove background noise from the audio data.

[0398] The server performs volume equalization to keep the audio volume level constant.

[0399] Input: Received audio data stored on the server

[0400] Output: Audio data with noise removed and volume leveled

[0401] Step 7: Extracting features from audio data

[0402] The server extracts features from the preprocessed speech data, including speaking rate, filler words, and stammers.

[0403] Input: Audio data with noise removed and volume leveled

[0404] Output: Audio data with extracted features

[0405] Step 8: Identify the speaker's habits

[0406] The server identifies speaker habits based on the extracted features, including rapid speech, frequent use of filler words, and stammering.

[0407] Input: Audio data with extracted features

[0408] Output: Data with speaker habits identified

[0409] Step 9: Emotion Recognition

[0410] The emotion engine recognizes the user's emotions in real time from the voice data, including joy, anger, sadness, surprise, etc.

[0411] Input: Audio data with extracted features and identified habits

[0412] Output: User's emotional information

[0413] Step 10: Emotional feedback

[0414] The emotion engine transmits the recognized emotion information to the server.

[0415] Input: User's emotional information

[0416] Output: Emotion information sent to the server

[0417] Step 11: Habit Elimination and Emotional Adjustments

[0418] Based on the identified habits and emotional information, the server generates a dataset for habit removal, which involves applying a generative AI model.

[0419] The server adjusts the speed of fast-paced speech, removes fillers, and adjusts speech speed and volume based on emotional information.

[0420] Input: Data with speaker habits identified and emotional feedback

[0421] Output: Corrected and adjusted audio data

[0422] Step 12: Regenerating spontaneous speech

[0423] The server recreates natural speech based on the corrected and adjusted voice data.

[0424] The server runs the regenerated audio data through a sound quality optimization engine to convert it into a clear, audible format.

[0425] Input: Corrected and adjusted audio data

[0426] Output: Regenerated and optimized audio data

[0427] Step 13: Sending the Regenerated Audio

[0428] The server transmits the regenerated voice data to the terminal in real time.

[0429] Input: Regenerated and optimized audio data

[0430] Output: Audio data sent to the device

[0431] Step 14: Playing back audio data

[0432] The terminal receives the regenerated voice data and plays it back to the user, so that the user hears modified, natural-sounding speech.

[0433] Input: Audio data sent to the device

[0434] Output: The audio played to the user

[0435] (Application example 2)

[0436] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0437] When interacting with customers in real-world situations, customer support staff are required to respond immediately to their customers' emotions and reactions, but they lack the technology to appropriately recognize the speaker's habits and emotions and adjust their speech in real time. This can make natural communication difficult, which can lead to a decrease in customer satisfaction.

[0438] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0439] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, means for recognizing emotions and adjusting a method for correcting the voice data based on the recognition results, and means for performing voice analysis and emotion recognition when interacting with a customer in real space and supporting a response according to the customer's emotions. This allows voice to be regenerated as natural speech in real time in accordance with the customer's emotions, enabling customer support staff to communicate more smoothly.

[0440] "Sound data" refers to an acoustic signal expressed in audio format.

[0441] "Input means" refers to a hardware or software mechanism for inputting voice data into the system.

[0442] The "analyzing means" is a component that has the function of processing input voice data and extracting specific features.

[0443] "Features" are data extracted from speech data that indicate the speaker's habits and speech characteristics.

[0444] The "means for identifying habits" is a device that has the function of identifying specific phrases, fillers, etc. from a speaker's speech.

[0445] The "means for removing" is a system that has the function of performing processing to remove the habits of the identified speaker.

[0446] The "regenerating means" is a mechanism for generating natural speech based on speech data from which habits have been removed.

[0447] A "means for playing" is a device for providing a user with reproduced natural speech sounds.

[0448] The "means for recognizing emotions" is a component that has the function of identifying emotions from the user's voice data and feeding that information back to the system.

[0449] The "means for adjusting the correction method" is a mechanism having a function for dynamically changing the method for correcting voice data based on the emotion recognition result.

[0450] "Real space" refers to the physical world, the environment in which a user actually exists and acts.

[0451] A "customer" is an entity that uses a service and is the person with whom the service is interacted.

[0452] "Voice analysis during dialogue" is the process of analyzing voice data in real time during a conversation with a customer and understanding its content.

[0453] The "means for emotion recognition" is a mechanism for analyzing and recognizing customer emotions in real time from voice data during a conversation.

[0454] The "means for supporting the response" is a device that has the function of supporting the customer support staff in order to respond appropriately according to the customer's feelings.

[0455] This invention is applied to a customer support system that handles customer support in the real world, and analyzes voice data in real time, removing the speaker's habits and regenerating natural speech. In addition, by using an emotion engine, the content of speech is adjusted based on the customer's emotions.

[0456] System Configuration

[0457] The system consists of the following main components:

[0458] Terminal: A device that inputs voice data and transmits the data to a server. As a concrete example, smart glasses are used.

[0459] Server: Analyzes voice data, extracts features, identifies and removes habits, reproduces natural speech, and recognizes emotions. This uses the Google Speech Recognition API and Transformers library.

[0460] User: The entity that inputs voice data through the system and receives modified, natural-sounding speech. Specifically, customer support staff.

[0461] Hardware configuration and data calculation

[0462] 1. Capture and transmit audio data

[0463] The device captures audio data through a microphone built into the smart glasses, which is temporarily buffered within the device and then transmitted to the server in real time.

[0464] 2. Voice Recognition

[0465] The server converts the voice data received from the device into text using the Google Speech Recognition API, which provides highly accurate voice recognition.

[0466] 3. Emotion recognition

[0467] The server uses an emotion engine (a model using the Transformers library) based on the converted text to identify the customer's emotions, which typically include emotion categories such as joy, anger, and sadness.

[0468] 4. Correcting speech habits

[0469] The server processes the text to identify and remove slurred speech, fillers, rapid speech, and other habits. The removed features are saved as the corrected text.

[0470] 5. Reproducing natural speech

[0471] The server then regenerates the corrected text and produces natural-sounding speech, adjusting the volume and speech rate based on the emotional information obtained from the emotion engine.

[0472] 6. Playing the corrected audio

[0473] The terminal receives the modified voice data sent from the server and plays it back to the user (customer support staff) through the smart glasses' speaker.

[0474] Specific examples

[0475] Physical store scenario

[0476] 1. Customer: "Um, can you tell me what features this product has?"

[0477] 2. The smart glasses capture the customer's speech and send the voice data to a PC.

[0478] 3. The PC receives the voice data and converts it into text using the Google Speech Recognition API.

[0479] 4. The server performs sentiment analysis on the converted text and recognizes the sentiment of "high interest."

[0480] 5. The program corrected the sentence by removing the "um" from "Um, can you tell me what features this product has?"

[0481] 6. Revised text: "Can you tell me what features this product has?"

[0482] 7. The regenerated audio is played back to the customer support staff through the smart glasses' speakers.

[0483] Prompt Sentence Examples

[0484] Question: "Please explain how the program can correct and regenerate speech when a customer says, 'Um, can you tell me what features this product has?'"

[0485] Example input: Customer: "Um, can you tell me what features this product has?"

[0486] Example output: Regenerated speech: "Can you tell me what features this product has?"

[0487] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0488] Step 1:

[0489] Capture and transmit audio data

[0490] A microphone built into the user's smart glasses captures the customer's speech in real time, generating digital audio data. This audio data is temporarily buffered inside the device and then sent to the server via the device's communication function. The input is the customer's speech, and the output is digital audio data sent to the server.

[0491] Step 2:

[0492] Voice Recognition

[0493] The server processes the voice data received from the device and converts it into text using the Google Speech Recognition API. This process results in the content of the voice data being expressed in text form. The input is the voice data sent to the server, and the output is the corresponding text data.

[0494] Step 3:

[0495] emotion recognition

[0496] The server analyzes emotions based on the converted text data using the Transformers library. It uses a generative AI model to analyze emotions from the customer's speech and obtains the results. The input is the converted text data, and the output is the customer's emotional data (e.g., joy, anger, sadness, etc.).

[0497] Step 4:

[0498] Correcting speech habits

[0499] The server identifies and removes speaker habits (fillers, slurred speech, rapid speech, etc.) from the text data along with the emotion recognition results. At this stage, unnecessary fillers and meaningless words are removed from the text. The inputs are the emotion recognition results and the text data, and the output is the corrected text data.

[0500] Step 5:

[0501] Natural speech reproduction

[0502] The server regenerates natural-sounding speech based on the corrected text data and emotion recognition results, adjusting the volume and speech rate. During this process, the text data is converted into clear speech data. The inputs are the corrected text data and emotion recognition results, and the output is the regenerated speech data.

[0503] Step 6:

[0504] Playing the corrected audio

[0505] The terminal receives the regenerated voice data sent from the server and plays it back to the user through the smart glasses speaker. In this process, the customer support staff as the user can hear the corrected natural-sounding speech. The input is the regenerated voice data, and the output is the reproduced voice provided to the user.

[0506] At each step, appropriate data processing or calculation is performed based on specific input data, and the output data required for the next step is generated as a result, thereby realizing natural communication with customers.

[0507] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0508] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0509] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0510] [Second embodiment]

[0511] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0512] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0513] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0514] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0515] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0516] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0517] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0518] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0519] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0520] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0521] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0522] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0523] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used with devices such as telephones, web conferences, and earphones with microphones. Specific embodiments of the present invention are described below.

[0524] System Configuration

[0525] The system consists of the following main components:

[0526] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[0527] Server: A processing device that receives voice data, analyzes it in real time, removes habits, and reproduces natural speech.

[0528] User: The entity that inputs speech and receives modified, natural-speech audio.

[0529] Program processing

[0530] The operation of the system will be specifically described below.

[0531] Input and transmission of voice data

[0532] 1. The user inputs audio using earphones with a microphone or a PC microphone. This audio is normal conversation.

[0533] 2. The device captures the audio data in real time and prepares to send it to the server.

[0534] 3. The device sends the captured audio data to the server.

[0535] Analyzing voice data and identifying habits

[0536] 4. The server receives the voice data sent from the device.

[0537] 5. The server performs preprocessing on the received audio data, including noise reduction and volume adjustment.

[0538] 6. The server extracts features from the audio data, including features such as rapid speech, stammers, and fillers.

[0539] 7. The server identifies the speaker's habits based on the extracted features.

[0540] Removes quirks and recreates natural speech

[0541] 8. The server applies generative AI models to remove identified habits, e.g., remove fillers and convert fast-paced parts to an appropriate speed.

[0542] 9. The server recreates natural-sounding speech based on the data after the habits have been removed.

[0543] Sending and playing back modified audio

[0544] 10. The server sends the regenerated audio data to the device.

[0545] 11. The device plays the modified audio data to the user, allowing the user to hear natural, fluent speech.

[0546] Specific examples

[0547] A specific example of use is shown below.

[0548] Call Scenarios

[0549] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0550] The device captures this audio and transmits it to the server in real time.

[0551] The server analyzes the audio data and identifies fillers such as "um" and "right."

[0552] The server removes the filler and converts it into natural speech, such as "I'd like to ask for a moment of your time."

[0553] The server sends the regenerated audio to the terminal, which plays it back.

[0554] The user can continue the conversation as if they were speaking fluently while listening to the other party's responses.

[0555] This invention enables users who are unsure about their speaking style or who stutter to communicate naturally and smoothly. This system will be an extremely useful tool in both business and personal settings.

[0556] The processing flow will be explained below.

[0557] Specific processing steps of the program

[0558] Step 1:

[0559] The user inputs audio using earphones with a microphone or the microphone on the PC.

[0560] The terminal captures the audio data in real time and stores the data in a buffer.

[0561] Step 2:

[0562] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0563] The device sends the audio data to the server in real time whenever possible.

[0564] Step 3:

[0565] The server receives the voice data transmitted from the terminal.

[0566] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[0567] Step 4:

[0568] The server adjusts the volume to set the optimal listening level.

[0569] The server completes preprocessing in order to analyze the audio data in real time.

[0570] Step 5:

[0571] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0572] The server analyzes the extracted features and identifies speaking habits.

[0573] Step 6:

[0574] The server generates a dataset to correct the identified habits.

[0575] The server applies a generative AI model to make corrections such as adjusting fast-talking parts and removing filler.

[0576] Step 7:

[0577] The server recreates natural speech based on the modified voice data.

[0578] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0579] Step 8:

[0580] The server transmits the regenerated voice data to the terminal in real time.

[0581] The terminal receives the modified natural speech audio and plays it back to the user.

[0582] Specific examples

[0583] Call Scenarios

[0584] Step 1:

[0585] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0586] The device captures this audio in real time and stores it in a buffer.

[0587] Step 2:

[0588] The device sends the captured audio data to the server.

[0589] Step 3:

[0590] The server receives the audio data and performs noise reduction.

[0591] Step 4:

[0592] The server adjusts the volume.

[0593] Step 5:

[0594] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0595] The server identifies the speech habits based on the identified feature amount.

[0596] Step 6:

[0597] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[0598] Step 7:

[0599] The server reproduces the natural utterance "Please give me a moment."

[0600] Step 8:

[0601] The server sends the regenerated audio to the terminal, which plays it back to the user.

[0602] Example 1

[0603] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0604] Conventional speech data analysis systems have had difficulty accurately removing speaker habits and regenerating natural speech. In particular, there is a need to provide natural conversations by removing characteristic habits such as rapid speech and fillers in real time. To solve this problem, an efficient and accurate method for analyzing and regenerating speech data is required.

[0605] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0606] In this invention, the server includes means for receiving speech data and performing preprocessing such as noise reduction and volume adjustment, means for extracting features from the speech data and identifying the speaker's habits, and means for applying a generative model to remove the identified habits and regenerate natural speech. This makes it possible to generate natural speech in real time while removing the speaker's unnatural habits.

[0607] "Voice data" refers to voice information stored in digital format that a user makes through an input device such as a microphone.

[0608] "Real-time" refers to data and events being processed as they occur, with immediate results available without delay.

[0609] "Noise reduction" is a technology for improving the clarity of audio by removing unnecessary background sounds and noise from audio data.

[0610] "Volume adjustment" is a process for equalizing the volume level of recorded audio data to ensure that it is easy to listen to when played back.

[0611] "Features" refer to specific patterns or attributes extracted from speech data, and are information used to identify a speaker's habits and characteristics.

[0612] A "generative model" is an artificial intelligence model used to create new data from existing data, and in this case is specifically used to regenerate audio data.

[0613] "Fillers" refer to unnecessary parts of speech, such as "ums" and "hmms," that speakers unconsciously use to connect words.

[0614] "Habits" refer to a speaker's particular speech patterns or habitual speaking characteristics that need to be identified and eliminated.

[0615] "Regeneration" refers to the process of reconstructing processed data in its original or new form, which in this case means producing natural-sounding speech.

[0616] This system analyzes input speech data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used on devices such as telephones, web conferences, and earphones with microphones.

[0617] System Configuration

[0618] The system consists of the following main components:

[0619] Terminal: A device that allows a user to input voice and transmit the voice data to a server. Examples include smartphones, PCs, and tablets.

[0620] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and regenerates natural speech. The server is equipped with a high-performance processor and sufficient memory.

[0621] User: The entity that inputs speech and receives modified, natural-speech audio.

[0622] Hardware and Software

[0623] The following hardware and software are used to implement this system.

[0624] Hardware:

[0625] Earphones with a microphone or a microphone built into your PC: Captures your voice.

[0626] Devices such as smartphones, PCs, and tablets: Captures audio data in real time and sends it to a server.

[0627] Server equipped with a high-performance processor: Analyzes and processes voice data.

[0628] software:

[0629] Audio processing library "SoX": Performs preprocessing for noise reduction and volume adjustment.

[0630] Machine learning library "TensorFlow": Extracts features from audio data and identifies habits.

[0631] Generative AI model "GPT-3": Removes fillers and reproduces natural speech.

[0632] "Vosk" speech generation library: converts text data into acoustic data.

[0633] Specific examples

[0634] Below are some specific examples of how the system can be used.

[0635] Call Scenarios

[0636] 1. During a call, the user says, "Um, I'd like to have a moment of your time."

[0637] 2. The device captures this audio and sends it to the server in real time.

[0638] 3. The server analyzes the audio data and identifies fillers such as "uh" and "right."

[0639] 4. The server removes the filler and converts the utterance into natural speech: "Please give me a moment of your time."

[0640] 5. The server sends the regenerated audio to the device, which plays it back.

[0641] 6. Users can continue the conversation as if they were speaking fluently while listening to the responses of the other party.

[0642] Prompt Sentence Examples

[0643] Below are some examples of prompts that can be given to the generative AI model.

[0644] Prompt Sentence Examples

[0645] Input: "Um, I'd like to, well, take a moment."

[0646] Instructions to the generative AI model: "Remove fillers and convert to natural-sounding speech."

[0647] In this way, a system is provided that allows the speaker to have a smooth conversation.

[0648] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0649] Step 1:

[0650] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is a normal conversation, such as "Um, I'd like to have a moment of your time." The input voice data is captured as an analog signal and converted to a digital signal within the device.

[0651] Step 2:

[0652] The device stores the captured audio data in memory in real time and prepares it for transmission to the server. During this process, the audio data is divided into packets at short time intervals (for example, every second). The data is then sent to the server using a secure communication protocol such as HTTPS. The input is analog audio data, and the output is packetized digital audio data.

[0653] Step 3:

[0654] The server receives the audio data sent from the terminal. The received packets are reassembled to form a continuous audio data stream. The input is packetized digital audio data, and the output is a reassembled continuous audio data stream.

[0655] Step 4:

[0656] The server performs pre-processing for noise reduction and volume adjustment on the received audio data. This process uses the "SoX" library to remove background noise and equalize the volume. Specifically, it applies noise filtering and audio normalization algorithms. The input is a reconstructed continuous audio data stream, and the output is pre-processed, clean audio data.

[0657] Step 5:

[0658] The server extracts features from the preprocessed audio data. In this step, the machine learning library "TensorFlow" is used to calculate features such as Mel-Frequency Cepstrum Coefficients (MFCCs) from the audio data. The features represent specific attributes of the audio data and are used as input data to identify the speaker's habits. The input is preprocessed, clean audio data, and the output is the extracted audio features.

[0659] Step 6:

[0660] The server identifies the speaker's habits based on the extracted features. This process uses an algorithm trained on previous datasets to detect speech habits such as quick speech, stammers, and fillers. The input is the speech features, and the output is the identified speaker's habits.

[0661] Step 7:

[0662] The server applies a generative model to remove the identified habits. Specifically, it uses the GPT-3 model and provides a prompt, such as "Please remove fillers and convert to natural speech." The generative model removes unnecessary fillers and habits and converts the speech to natural speech. The input is speaker habit information and audio data, and the output is natural speech data with the habits removed.

[0663] Step 8:

[0664] The server recreates natural speech data with the quirks removed. In this step, the "Vosk" library is used to convert text data into acoustic data and generate natural speech. The input is the identified audio-text data, and the output is the recreated continuous audio data.

[0665] Step 9:

[0666] The server sends the regenerated audio data to the device, where it is repacketized and transmitted securely via a protocol such as HTTPS. The input is the regenerated continuous audio data, and the output is packetized digital audio data.

[0667] Step 10:

[0668] The terminal receives the modified audio data and plays it back to the user. Specifically, it reconstructs the received packetized data and plays it back to the user through a speaker or earphone. The input is packetized digital audio data, and the output is reproduced audio data that the user can hear.

[0669] (Application example 1)

[0670] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0671] In a typical store, if a store clerk has a habit of speaking in a certain way when interacting with customers, it can cause discomfort to the customer. In particular, speaking too quickly, stammering, and using too many fillers can be obstacles to smooth communication with customers. Unless this issue is resolved, there is a risk that the quality of store service will decline and customer satisfaction will decrease. Therefore, the present invention aims to provide a system that automatically removes the habit of store clerks' speaking in a certain way and converts it into natural and fluent speech, thereby facilitating communication between store clerks and customers and improving customer satisfaction.

[0672] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0673] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, and means for analyzing and correcting the voice of the store clerk to smoothly serve customers in the store. This enables the store clerk to speak naturally and fluently without worrying about their speaking habits.

[0674] "Audio data" refers to data in which audio information is recorded in digital format.

[0675] "Analyze in real time" means processing data as it is entered.

[0676] "Speaker features" are data that represent specific patterns and attributes contained in the speaker's voice.

[0677] "De-embedding" means removing or modifying unwanted features in a speaker's speech (e.g., tongue twisters, fillers, etc.).

[0678] "Natural speech" refers to fluent and smooth speech that does not sound strange to the listener.

[0679] "Playback" means outputting the processed audio data as sound through speakers or headphones.

[0680] A "system" is a structure that includes multiple components that work in conjunction with each other.

[0681] A "store" is a physical location for selling goods and services.

[0682] "Customer service" means the service and interaction activities that store staff engage in with customers.

[0683] "Modification" means to change the original state and adjust it to a desired form.

[0684] The present invention is a system that analyzes voice data of store clerks in real time, automatically removes speaking habits, and reproduces natural speech in order to facilitate smooth customer service in stores. This system uses a device such as a smartphone or earphones with a microphone to capture the voice of the store clerk, and processes the voice data on a server. Specific embodiments of the present invention are described below.

[0685] System Configuration

[0686] The system consists of the following main components:

[0687] Terminal: A device that allows the store clerk to input voice and transmit the voice data to the server. Specifically, a smartphone or earphones with a microphone are used.

[0688] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and reproduces natural speech.

[0689] Store Clerk: The subject who inputs speech and receives modified natural speech audio.

[0690] Program Processing Overview

[0691] The program of this system executes the following processes.

[0692] 1. The terminal captures the clerk's voice in real time and sends it to the server.

[0693] 2. The server receives the audio data and performs preprocessing such as noise reduction and volume adjustment.

[0694] 3. The server extracts features from the speech data, including features such as rapid speech, stammers, and fillers.

[0695] 4. The server identifies the speaker's habits based on the extracted features.

[0696] 5. The server applies generative AI models to remove quirks, such as removing fillers and converting fast-paced parts to the appropriate speed.

[0697] 6. The server recreates natural-sounding speech based on the data after the habits have been removed.

[0698] 7. The server sends the regenerated audio data to the device.

[0699] 8. The terminal plays the corrected voice data to the store clerk.

[0700] Hardware used

[0701] Smartphone: A device for voice input and playback.

[0702] Earphones with microphone: A device for high-quality audio capture.

[0703] Software used

[0704] Python: The base programming language.

[0705] SpeechRecognition: A library for speech capture and recognition.

[0706] requests: An HTTP client library for sending audio data to a server.

[0707] Specific examples

[0708] For example, if a store clerk says, "Um, well, I'd like to have a moment of your time," the device captures this speech and sends it to the server in real time. The server analyzes the speech data, identifies fillers such as "um" and "I see," and removes them to convert it into natural speech, "I'd like to have a moment of your time." The server then sends the regenerated speech data to the device, which plays it back to the store clerk. This allows the store clerk to interact with the customer as if they were speaking fluently.

[0709] Prompt Sentence Examples

[0710] For the generative AI model that "removes fillers" and "converts to natural speech," use prompts like these:

[0711] It analyzes the voice data, identifies and removes unnecessary fillers (such as "um" or "hmm"), converts the speech into natural-sounding speech, and converts fast-paced speech to an appropriate speed, outputting smooth speech.

[0712] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0713] Step 1:

[0714] The user inputs voice using earphones with a microphone or the microphone on a smartphone. At this time, the user conducts normal conversation and the voice data is captured. The input voice is stored as raw data on the device.

[0715] Step 2:

[0716] The device transmits the captured audio data to the server in real time. Specifically, the device converts the audio data into a digital format and sends the data to the server using an HTTP request, with the audio data remaining raw.

[0717] Step 3:

[0718] The server receives the voice data sent from the device, converts the received data into an analyzable format, and uses it in the next processing step.

[0719] Step 4:

[0720] The server preprocesses the audio data. Specifically, it performs noise reduction and volume adjustment. This processing improves the quality of the audio data, making it easier to perform subsequent analysis. The input is raw audio data, and the output is audio data with noise reduced and volume adjusted.

[0721] Step 5:

[0722] The server extracts speaker features from the audio data. For example, it uses a speech recognition library to identify features such as rapid speech, stammers, and fillers from the audio signal. At this stage, features are generated as numerical data or in the form of tags.

[0723] Step 6:

[0724] The server identifies the speaker's habits based on the extracted features. For example, if there are many fillers, it will tag it as a "filler habit," and if the speaker speaks quickly, it will identify it as a "fast speaking habit." In this step, the habit data is output as the result of the feature analysis.

[0725] Step 7:

[0726] The server applies a generative AI model to remove the identified habits. Specifically, it modifies the voice data using prompts for the generative AI model ("Analyze the voice data, identify and remove unnecessary fillers (such as 'um' or 'hmm'), and convert it into natural speech. Convert fast-paced parts to an appropriate speed and output smooth speech."). The input is the voice data with identified habits, and the output is the modified natural speech data.

[0727] Step 8:

[0728] The server sends the voice data regenerated into natural speech to the device. The corrected voice data is sent back to the device using an HTTP request. The input is the natural speech data, and the output is the completion status of transmission to the device.

[0729] Step 9:

[0730] The terminal plays the corrected voice data to the user. The clerk can hear fluent and natural speech through a playback device (earphones or smartphone speakers). The input is the corrected voice data received from the server, and the output is the voice that the user can hear.

[0731] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0732] The present invention relates to a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. The system further includes an emotion engine that recognizes the user's emotions and adjusts the method of modifying the voice data based on the emotions.

[0733] System Configuration

[0734] The system consists of the following main components:

[0735] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[0736] Server: A processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech.

[0737] User: The entity that inputs speech and receives modified, natural-speech audio.

[0738] Emotion engine: A module that recognizes the user's emotions in real time and feeds the analysis results back to the system.

[0739] Program processing

[0740] The specific processing of the system will be explained below.

[0741] Input and transmission of voice data

[0742] The user inputs voice using earphones with a microphone or a PC microphone. This voice is normal conversation.

[0743] The terminal captures the audio data in real time and stores the data in a buffer.

[0744] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0745] The device sends the audio data to the server in real time whenever possible.

[0746] Voice data analysis and emotion recognition

[0747] The server receives the voice data transmitted from the terminal.

[0748] The server performs pre-processing of the received audio data, including noise reduction and volume adjustment.

[0749] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0750] The server analyzes the extracted features and identifies speaking habits.

[0751] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[0752] Habit removal and emotional adjustment

[0753] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[0754] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0755] Reproduction and playback of spontaneous speech

[0756] The server recreates natural speech based on the modified voice data.

[0757] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0758] The server transmits the regenerated voice data to the terminal in real time.

[0759] The terminal receives the modified natural speech audio and plays it back to the user.

[0760] Specific examples

[0761] Call Scenarios

[0762] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0763] The device captures this audio in real time and stores it in a buffer.

[0764] The device sends the captured audio data to the server.

[0765] The server receives the audio data and performs noise reduction and volume adjustment.

[0766] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0767] The server identifies the speech habits based on the identified feature amount.

[0768] The emotion engine recognizes that the user is feeling a little nervous and sends that information to the server.

[0769] The server will remove identified fillers such as "uh" and "right," convert fast-paced parts to an appropriate speed, and adjust the volume slightly to provide a sense of security.

[0770] The server reproduces the natural utterance "Please give me a moment."

[0771] The server sends the regenerated audio to the terminal, which plays it back to the user.

[0772] This invention enables natural and smooth communication even for users who are unsure about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses are realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[0773] The processing flow will be explained below.

[0774] Specific processing steps of the program

[0775] Step 1:

[0776] The user inputs audio using earphones with a microphone or the microphone on the PC.

[0777] The terminal captures the audio data in real time and stores the data in a buffer.

[0778] Step 2:

[0779] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[0780] The device sends the audio data to the server in real time whenever possible.

[0781] Step 3:

[0782] The server receives the voice data transmitted from the terminal.

[0783] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[0784] The server adjusts the volume to set the optimal listening level.

[0785] Step 4:

[0786] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[0787] The server analyzes the extracted features and identifies speaking habits.

[0788] Step 5:

[0789] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[0790] The emotion engine feeds back the recognized emotion information to the server.

[0791] Step 6:

[0792] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[0793] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0794] Step 7:

[0795] The server reproduces natural-spoken speech with the quirks removed and the emotions reflected.

[0796] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[0797] Step 8:

[0798] The server transmits the regenerated voice data to the terminal in real time.

[0799] The terminal receives the modified natural speech audio and plays it back to the user.

[0800] Specific examples

[0801] Call Scenarios

[0802] Step 1:

[0803] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[0804] The device captures this audio in real time and stores it in a buffer.

[0805] Step 2:

[0806] The device sends the captured audio data to the server.

[0807] Step 3:

[0808] The server receives the audio data and performs noise reduction and volume adjustment.

[0809] Step 4:

[0810] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[0811] The server identifies the speech habits based on the identified feature amount.

[0812] Step 5:

[0813] The emotion engine recognizes that the user is feeling a little nervous.

[0814] The emotion engine sends the recognized emotion information to the server.

[0815] Step 6:

[0816] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[0817] Based on feedback from the emotion engine, the server makes adjustments such as slowing down the speaking rate and lowering the volume.

[0818] Step 7:

[0819] The server reproduces the natural utterance "Please give me a moment."

[0820] The server converts the reproduced audio data into a clear, audible format.

[0821] Step 8:

[0822] The server transmits the regenerated voice data to the terminal.

[0823] The terminal plays the modified natural speech to the user.

[0824] Example 2

[0825] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0826] Conventional speech recognition systems have had the problem of being unable to completely remove unnatural speech caused by the speaker's unique habits and emotions. Furthermore, they were unable to recognize the speaker's emotions and adjust their speech accordingly, making it difficult to achieve more natural and personalized speech communication. This has made it difficult for users who are unsure about their speaking style or who stutter to communicate smoothly.

[0827] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0828] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time, means for extracting features from the analyzed voice data, means for identifying the speaker's habits based on the extracted features and removing the habits, means for recognizing the user's emotions in real time and sending the information to the analysis means, means for regenerating the voice data from which the habits have been removed and which has been adjusted based on the emotion information as natural speech, and means for playing back the regenerated voice data to the user. This makes it possible to adjust and optimize the voice data based on the speaker's habits and emotions, thereby achieving natural and smooth communication.

[0829] "Means for inputting voice data" refers to equipment or a method by which a user inputs voice data into the system using a microphone or other input device.

[0830] "Means for real-time analysis" refers to algorithms or software that process and analyze input voice data almost instantaneously.

[0831] "Means for extracting features" refers to a process for extracting specific patterns or attributes (e.g., speaking rate, fillers, stammers, etc.) from speech data.

[0832] "Means for identifying and removing speaker habits" refers to technology that detects a speaker's specific habits (e.g., speaking quickly, using fillers, etc.) based on extracted features and corrects or removes them.

[0833] The "means for recognizing the user's emotions in real time and transmitting that information to the analysis means" is a module for identifying the user's emotional state (e.g., joy, anger, sadness, surprise, etc.) from voice data in real time and reflecting that information in voice analysis.

[0834] A "regenerating means" is an algorithm or device that uses the modified speech data to reconstruct natural speech.

[0835] The "playback means" refers to an audio output device such as a speaker or earphones that allows the user to hear the regenerated audio data.

[0836] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, regenerates natural speech, and further recognizes the user's emotions and corrects the voice data based on those emotions. An embodiment of the present invention consists of the following main components and processing flow.

[0837] System Configuration

[0838] 1. Terminal

[0839] A device that allows a user to input voice and transmit the voice data to a server. Specific examples include earphones with a microphone and a microphone connected to a PC.

[0840] 2. Server

[0841] It is a processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech. The server processes the voice data using an advanced generative AI model.

[0842] 3. Emotion Engine

[0843] This module recognizes the user's emotions in real time and feeds the analysis results back to the system. The emotion engine extracts emotions such as joy, anger, sadness, and surprise from voice data.

[0844] 4. Users

[0845] A subject that inputs speech and receives modified, natural-speech audio.

[0846] Program processing

[0847] Input and transmission of voice data

[0848] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is captured in real time by the device and stored in a buffer. The device then prepares the voice data for transmission and starts sending it to the server. Transmission is performed as quickly as possible in real time.

[0849] Receiving and preprocessing audio data

[0850] The server receives the audio data sent from the device, applies a noise reduction filter to the received audio data to remove background noise, and performs volume equalization to adjust and maintain a constant volume level.

[0851] Voice data analysis and emotion recognition

[0852] The server extracts features from the preprocessed voice data. These features include speaking rate, fillers (e.g., "um," "ah," etc.), and stammers. The server analyzes the features of the voice data and identifies speaking habits. At the same time, the emotion engine recognizes the user's emotions from the voice data in real time and sends that information to the server.

[0853] Habit removal and emotional adjustment

[0854] The server generates a dataset for removing habits based on the identified habits and emotional information. Specifically, it uses a generative AI model to modify the speech data, removing identified fillers and adjusting fast-paced parts to an appropriate speed. It also adjusts speech rate and volume based on feedback from the emotion engine.

[0855] Reproduction and playback of spontaneous speech

[0856] The server uses the modified voice data to recreate natural-sounding speech. The recreated voice data is passed through a sound quality optimization engine, which converts it into a clear, easy-to-listen format. Finally, the server transmits the optimized voice data to the device in real time, and the device plays the received voice back to the user.

[0857] Specific examples

[0858] Call Scenarios

[0859] During a call, a user utters, "Um, well, I'd like to have a moment of your time." The device captures this audio in real time and stores it in a buffer. The device then sends the stored audio data to the server. The server receives the audio data and performs noise reduction and volume adjustment. Features such as speech rate, fillers, and stammers are extracted from the audio data and speech habits are identified based on these. At the same time, the emotion engine recognizes whether the user is feeling nervous and sends this information to the server. The server then removes the identified fillers, adjusts the fast-talking parts to an appropriate speed, and adjusts the volume to provide a sense of security. The regenerated, natural utterance, "I'd like to have a moment of your time," is then sent to the device, which plays it back to the user.

[0860] This system enables natural and smooth communication even for users who are anxious about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses can be realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[0861] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0862] Step 1: Input audio data

[0863] The user inputs voice using earphones with a microphone or the microphone on the PC. At this stage, the user can have a normal conversation.

[0864] Input: User speaking

[0865] Output: Analog audio signal

[0866] Step 2: Capturing and buffering audio data

[0867] The terminal captures the user's input voice in real time and converts it into a digital voice signal.

[0868] The device temporarily stores captured digital audio in a buffer.

[0869] Input: Analog audio signal

[0870] Output: Digital audio data

[0871] Step 3: Prepare to send audio data

[0872] The terminal reads the voice data stored in the buffer at regular intervals and prepares it for transmission, performing preprocessing such as data compression.

[0873] Input: Buffered digital audio data

[0874] Output: Audio data format that can be sent to the server

[0875] Step 4: Sending audio data

[0876] The data is converted into a voice data format that can be transmitted by the terminal and sent to the server. Transmission is performed as quickly as possible in real time.

[0877] Input: Audio data format that can be sent to the server

[0878] Output: Audio data sent to the server

[0879] Step 5: Receiving audio data

[0880] The server receives the voice data transmitted from the terminal in real time.

[0881] Input: Audio data sent from the device

[0882] Output: Received audio data stored on the server

[0883] Step 6: Preprocessing the audio data

[0884] The server applies a noise reduction filter to remove background noise from the audio data.

[0885] The server performs volume equalization to keep the audio volume level constant.

[0886] Input: Received audio data stored on the server

[0887] Output: Audio data with noise removed and volume leveled

[0888] Step 7: Extracting features from audio data

[0889] The server extracts features from the preprocessed speech data, including speaking rate, filler words, and stammers.

[0890] Input: Audio data with noise removed and volume leveled

[0891] Output: Audio data with extracted features

[0892] Step 8: Identify the speaker's habits

[0893] The server identifies speaker habits based on the extracted features, including rapid speech, frequent use of filler words, and stammering.

[0894] Input: Audio data with extracted features

[0895] Output: Data with speaker habits identified

[0896] Step 9: Emotion Recognition

[0897] The emotion engine recognizes the user's emotions in real time from the voice data, including joy, anger, sadness, surprise, etc.

[0898] Input: Audio data with extracted features and identified habits

[0899] Output: User's emotional information

[0900] Step 10: Emotional feedback

[0901] The emotion engine transmits the recognized emotion information to the server.

[0902] Input: User's emotional information

[0903] Output: Emotion information sent to the server

[0904] Step 11: Habit Elimination and Emotional Adjustments

[0905] Based on the identified habits and emotional information, the server generates a dataset for habit removal, which involves applying a generative AI model.

[0906] The server adjusts the speed of fast-paced speech, removes fillers, and adjusts speech speed and volume based on emotional information.

[0907] Input: Data with speaker habits identified and emotional feedback

[0908] Output: Corrected and adjusted audio data

[0909] Step 12: Regenerating spontaneous speech

[0910] The server recreates natural speech based on the corrected and adjusted voice data.

[0911] The server runs the regenerated audio data through a sound quality optimization engine to convert it into a clear, audible format.

[0912] Input: Corrected and adjusted audio data

[0913] Output: Regenerated and optimized audio data

[0914] Step 13: Sending the Regenerated Audio

[0915] The server transmits the regenerated voice data to the terminal in real time.

[0916] Input: Regenerated and optimized audio data

[0917] Output: Audio data sent to the device

[0918] Step 14: Playing back audio data

[0919] The terminal receives the regenerated voice data and plays it back to the user, so that the user hears modified, natural-sounding speech.

[0920] Input: Audio data sent to the device

[0921] Output: The audio played to the user

[0922] (Application example 2)

[0923] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0924] When interacting with customers in real-world situations, customer support staff are required to respond immediately to their customers' emotions and reactions, but they lack the technology to appropriately recognize the speaker's habits and emotions and adjust their speech in real time. This can make natural communication difficult, which can lead to a decrease in customer satisfaction.

[0925] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0926] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, means for recognizing emotions and adjusting a method for correcting the voice data based on the recognition results, and means for performing voice analysis and emotion recognition when interacting with a customer in real space and supporting a response according to the customer's emotions. This allows voice to be regenerated as natural speech in real time in accordance with the customer's emotions, enabling customer support staff to communicate more smoothly.

[0927] "Sound data" refers to an acoustic signal expressed in audio format.

[0928] "Input means" refers to a hardware or software mechanism for inputting voice data into the system.

[0929] The "analyzing means" is a component that has the function of processing input voice data and extracting specific features.

[0930] "Features" are data extracted from speech data that indicate the speaker's habits and speech characteristics.

[0931] The "means for identifying habits" is a device that has the function of identifying specific phrases, fillers, etc. from a speaker's speech.

[0932] The "means for removing" is a system that has the function of performing processing to remove the habits of the identified speaker.

[0933] The "regenerating means" is a mechanism for generating natural speech based on speech data from which habits have been removed.

[0934] A "means for playing" is a device for providing a user with reproduced natural speech sounds.

[0935] The "means for recognizing emotions" is a component that has the function of identifying emotions from the user's voice data and feeding that information back to the system.

[0936] The "means for adjusting the correction method" is a mechanism having a function for dynamically changing the method for correcting voice data based on the emotion recognition result.

[0937] "Real space" refers to the physical world, the environment in which a user actually exists and acts.

[0938] A "customer" is an entity that uses a service and is the person with whom the service is interacted.

[0939] "Voice analysis during dialogue" is the process of analyzing voice data in real time during a conversation with a customer and understanding its content.

[0940] The "means for emotion recognition" is a mechanism for analyzing and recognizing customer emotions in real time from voice data during a conversation.

[0941] The "means for supporting the response" is a device that has the function of supporting the customer support staff in order to respond appropriately according to the customer's feelings.

[0942] This invention is applied to a customer support system that handles customer support in the real world, and analyzes voice data in real time, removing the speaker's habits and regenerating natural speech. In addition, by using an emotion engine, the content of speech is adjusted based on the customer's emotions.

[0943] System Configuration

[0944] The system consists of the following main components:

[0945] Terminal: A device that inputs voice data and transmits the data to a server. As a concrete example, smart glasses are used.

[0946] Server: Analyzes voice data, extracts features, identifies and removes habits, reproduces natural speech, and recognizes emotions. This uses the Google Speech Recognition API and Transformers library.

[0947] User: The entity that inputs voice data through the system and receives modified, natural-sounding speech. Specifically, customer support staff.

[0948] Hardware configuration and data calculation

[0949] 1. Capture and transmit audio data

[0950] The device captures audio data through a microphone built into the smart glasses, which is temporarily buffered within the device and then transmitted to the server in real time.

[0951] 2. Voice Recognition

[0952] The server converts the voice data received from the device into text using the Google Speech Recognition API, which provides highly accurate voice recognition.

[0953] 3. Emotion recognition

[0954] The server uses an emotion engine (a model using the Transformers library) based on the converted text to identify the customer's emotions, which typically include emotion categories such as joy, anger, and sadness.

[0955] 4. Correcting speech habits

[0956] The server processes the text to identify and remove slurred speech, fillers, rapid speech, and other habits. The removed features are saved as the corrected text.

[0957] 5. Reproducing natural speech

[0958] The server then regenerates the corrected text and produces natural-sounding speech, adjusting the volume and speech rate based on the emotional information obtained from the emotion engine.

[0959] 6. Playing the corrected audio

[0960] The terminal receives the modified voice data sent from the server and plays it back to the user (customer support staff) through the smart glasses' speaker.

[0961] Specific examples

[0962] Physical store scenario

[0963] 1. Customer: "Um, can you tell me what features this product has?"

[0964] 2. The smart glasses capture the customer's speech and send the voice data to a PC.

[0965] 3. The PC receives the voice data and converts it into text using the Google Speech Recognition API.

[0966] 4. The server performs sentiment analysis on the converted text and recognizes the sentiment of "high interest."

[0967] 5. The program corrected the sentence by removing the "um" from "Um, can you tell me what features this product has?"

[0968] 6. Revised text: "Can you tell me what features this product has?"

[0969] 7. The regenerated audio is played back to the customer support staff through the smart glasses' speakers.

[0970] Prompt Sentence Examples

[0971] Question: "Please explain how the program can correct and regenerate speech when a customer says, 'Um, can you tell me what features this product has?'"

[0972] Example input: Customer: "Um, can you tell me what features this product has?"

[0973] Example output: Regenerated speech: "Can you tell me what features this product has?"

[0974] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0975] Step 1:

[0976] Capture and transmit audio data

[0977] A microphone built into the user's smart glasses captures the customer's speech in real time, generating digital audio data. This audio data is temporarily buffered inside the device and then sent to the server via the device's communication function. The input is the customer's speech, and the output is digital audio data sent to the server.

[0978] Step 2:

[0979] Voice Recognition

[0980] The server processes the voice data received from the device and converts it into text using the Google Speech Recognition API. This process results in the content of the voice data being expressed in text form. The input is the voice data sent to the server, and the output is the corresponding text data.

[0981] Step 3:

[0982] emotion recognition

[0983] The server analyzes emotions based on the converted text data using the Transformers library. It uses a generative AI model to analyze emotions from the customer's speech and obtains the results. The input is the converted text data, and the output is the customer's emotional data (e.g., joy, anger, sadness, etc.).

[0984] Step 4:

[0985] Correcting speech habits

[0986] The server identifies and removes speaker habits (fillers, slurred speech, rapid speech, etc.) from the text data along with the emotion recognition results. At this stage, unnecessary fillers and meaningless words are removed from the text. The inputs are the emotion recognition results and the text data, and the output is the corrected text data.

[0987] Step 5:

[0988] Natural speech reproduction

[0989] The server regenerates natural-sounding speech based on the corrected text data and emotion recognition results, adjusting the volume and speech rate. During this process, the text data is converted into clear speech data. The inputs are the corrected text data and emotion recognition results, and the output is the regenerated speech data.

[0990] Step 6:

[0991] Playing the corrected audio

[0992] The terminal receives the regenerated voice data sent from the server and plays it back to the user through the smart glasses speaker. In this process, the customer support staff as the user can hear the corrected natural-sounding speech. The input is the regenerated voice data, and the output is the reproduced voice provided to the user.

[0993] At each step, appropriate data processing or calculation is performed based on specific input data, and the output data required for the next step is generated as a result, thereby realizing natural communication with customers.

[0994] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0995] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0996] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0997] [Third embodiment]

[0998] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0999] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1000] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1001] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1002] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1003] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1004] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1005] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1006] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1007] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1008] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1009] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[1010] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used with devices such as telephones, web conferences, and earphones with microphones. Specific embodiments of the present invention are described below.

[1011] System Configuration

[1012] The system consists of the following main components:

[1013] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[1014] Server: A processing device that receives voice data, analyzes it in real time, removes habits, and reproduces natural speech.

[1015] User: The entity that inputs speech and receives modified, natural-speech audio.

[1016] Program processing

[1017] The operation of the system will be specifically described below.

[1018] Input and transmission of voice data

[1019] 1. The user inputs audio using earphones with a microphone or a PC microphone. This audio is normal conversation.

[1020] 2. The device captures the audio data in real time and prepares to send it to the server.

[1021] 3. The device sends the captured audio data to the server.

[1022] Analyzing voice data and identifying habits

[1023] 4. The server receives the voice data sent from the device.

[1024] 5. The server performs preprocessing on the received audio data, including noise reduction and volume adjustment.

[1025] 6. The server extracts features from the audio data, including features such as rapid speech, stammers, and fillers.

[1026] 7. The server identifies the speaker's habits based on the extracted features.

[1027] Removes quirks and recreates natural speech

[1028] 8. The server applies generative AI models to remove identified habits, e.g., remove fillers and convert fast-paced parts to an appropriate speed.

[1029] 9. The server recreates natural-sounding speech based on the data after the habits have been removed.

[1030] Sending and playing back modified audio

[1031] 10. The server sends the regenerated audio data to the device.

[1032] 11. The device plays the modified audio data to the user, allowing the user to hear natural, fluent speech.

[1033] Specific examples

[1034] A specific example of use is shown below.

[1035] Call Scenarios

[1036] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1037] The device captures this audio and transmits it to the server in real time.

[1038] The server analyzes the audio data and identifies fillers such as "um" and "right."

[1039] The server removes the filler and converts it into natural speech, such as "I'd like to ask for a moment of your time."

[1040] The server sends the regenerated audio to the terminal, which plays it back.

[1041] The user can continue the conversation as if they were speaking fluently while listening to the other party's responses.

[1042] This invention enables users who are unsure about their speaking style or who stutter to communicate naturally and smoothly. This system will be an extremely useful tool in both business and personal settings.

[1043] The processing flow will be explained below.

[1044] Specific processing steps of the program

[1045] Step 1:

[1046] The user inputs audio using earphones with a microphone or the microphone on the PC.

[1047] The terminal captures the audio data in real time and stores the data in a buffer.

[1048] Step 2:

[1049] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1050] The device sends the audio data to the server in real time whenever possible.

[1051] Step 3:

[1052] The server receives the voice data transmitted from the terminal.

[1053] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[1054] Step 4:

[1055] The server adjusts the volume to set the optimal listening level.

[1056] The server completes preprocessing in order to analyze the audio data in real time.

[1057] Step 5:

[1058] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1059] The server analyzes the extracted features and identifies speaking habits.

[1060] Step 6:

[1061] The server generates a dataset to correct the identified habits.

[1062] The server applies a generative AI model to make corrections such as adjusting fast-talking parts and removing filler.

[1063] Step 7:

[1064] The server recreates natural speech based on the modified voice data.

[1065] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1066] Step 8:

[1067] The server transmits the regenerated voice data to the terminal in real time.

[1068] The terminal receives the modified natural speech audio and plays it back to the user.

[1069] Specific examples

[1070] Call Scenarios

[1071] Step 1:

[1072] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1073] The device captures this audio in real time and stores it in a buffer.

[1074] Step 2:

[1075] The device sends the captured audio data to the server.

[1076] Step 3:

[1077] The server receives the audio data and performs noise reduction.

[1078] Step 4:

[1079] The server adjusts the volume.

[1080] Step 5:

[1081] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1082] The server identifies the speech habits based on the identified feature amount.

[1083] Step 6:

[1084] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[1085] Step 7:

[1086] The server reproduces the natural utterance "Please give me a moment."

[1087] Step 8:

[1088] The server sends the regenerated audio to the terminal, which plays it back to the user.

[1089] Example 1

[1090] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1091] Conventional speech data analysis systems have had difficulty accurately removing speaker habits and regenerating natural speech. In particular, there is a need to provide natural conversations by removing characteristic habits such as rapid speech and fillers in real time. To solve this problem, an efficient and accurate method for analyzing and regenerating speech data is required.

[1092] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1093] In this invention, the server includes means for receiving speech data and performing preprocessing such as noise reduction and volume adjustment, means for extracting features from the speech data and identifying the speaker's habits, and means for applying a generative model to remove the identified habits and regenerate natural speech. This makes it possible to generate natural speech in real time while removing the speaker's unnatural habits.

[1094] "Voice data" refers to voice information stored in digital format that a user makes through an input device such as a microphone.

[1095] "Real-time" refers to data and events being processed as they occur, with immediate results available without delay.

[1096] "Noise reduction" is a technology for improving the clarity of audio by removing unnecessary background sounds and noise from audio data.

[1097] "Volume adjustment" is a process for equalizing the volume level of recorded audio data to ensure that it is easy to listen to when played back.

[1098] "Features" refer to specific patterns or attributes extracted from speech data, and are information used to identify a speaker's habits and characteristics.

[1099] A "generative model" is an artificial intelligence model used to create new data from existing data, and in this case is specifically used to regenerate audio data.

[1100] "Fillers" refer to unnecessary parts of speech, such as "ums" and "hmms," that speakers unconsciously use to connect words.

[1101] "Habits" refer to a speaker's particular speech patterns or habitual speaking characteristics that need to be identified and eliminated.

[1102] "Regeneration" refers to the process of reconstructing processed data in its original or new form, which in this case means producing natural-sounding speech.

[1103] This system analyzes input speech data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used on devices such as telephones, web conferences, and earphones with microphones.

[1104] System Configuration

[1105] The system consists of the following main components:

[1106] Terminal: A device that allows a user to input voice and transmit the voice data to a server. Examples include smartphones, PCs, and tablets.

[1107] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and regenerates natural speech. The server is equipped with a high-performance processor and sufficient memory.

[1108] User: The entity that inputs speech and receives modified, natural-speech audio.

[1109] Hardware and Software

[1110] The following hardware and software are used to implement this system.

[1111] Hardware:

[1112] Earphones with a microphone or a microphone built into your PC: Captures your voice.

[1113] Devices such as smartphones, PCs, and tablets: Captures audio data in real time and sends it to a server.

[1114] Server equipped with a high-performance processor: Analyzes and processes voice data.

[1115] software:

[1116] Audio processing library "SoX": Performs preprocessing for noise reduction and volume adjustment.

[1117] Machine learning library "TensorFlow": Extracts features from audio data and identifies habits.

[1118] Generative AI model "GPT-3": Removes fillers and reproduces natural speech.

[1119] "Vosk" speech generation library: converts text data into acoustic data.

[1120] Specific examples

[1121] Below are some specific examples of how the system can be used.

[1122] Call Scenarios

[1123] 1. During a call, the user says, "Um, I'd like to have a moment of your time."

[1124] 2. The device captures this audio and sends it to the server in real time.

[1125] 3. The server analyzes the audio data and identifies fillers such as "uh" and "right."

[1126] 4. The server removes the filler and converts the utterance into natural speech: "Please give me a moment of your time."

[1127] 5. The server sends the regenerated audio to the device, which plays it back.

[1128] 6. Users can continue the conversation as if they were speaking fluently while listening to the responses of the other party.

[1129] Prompt Sentence Examples

[1130] Below are some examples of prompts that can be given to the generative AI model.

[1131] Prompt Sentence Examples

[1132] Input: "Um, I'd like to, well, take a moment."

[1133] Instructions to the generative AI model: "Remove fillers and convert to natural-sounding speech."

[1134] In this way, a system is provided that allows the speaker to have a smooth conversation.

[1135] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1136] Step 1:

[1137] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is a normal conversation, such as "Um, I'd like to have a moment of your time." The input voice data is captured as an analog signal and converted to a digital signal within the device.

[1138] Step 2:

[1139] The device stores the captured audio data in memory in real time and prepares it for transmission to the server. During this process, the audio data is divided into packets at short time intervals (for example, every second). The data is then sent to the server using a secure communication protocol such as HTTPS. The input is analog audio data, and the output is packetized digital audio data.

[1140] Step 3:

[1141] The server receives the audio data sent from the terminal. The received packets are reassembled to form a continuous audio data stream. The input is packetized digital audio data, and the output is a reassembled continuous audio data stream.

[1142] Step 4:

[1143] The server performs pre-processing for noise reduction and volume adjustment on the received audio data. This process uses the "SoX" library to remove background noise and equalize the volume. Specifically, it applies noise filtering and audio normalization algorithms. The input is a reconstructed continuous audio data stream, and the output is pre-processed, clean audio data.

[1144] Step 5:

[1145] The server extracts features from the preprocessed audio data. In this step, the machine learning library "TensorFlow" is used to calculate features such as Mel-Frequency Cepstrum Coefficients (MFCCs) from the audio data. The features represent specific attributes of the audio data and are used as input data to identify the speaker's habits. The input is preprocessed, clean audio data, and the output is the extracted audio features.

[1146] Step 6:

[1147] The server identifies the speaker's habits based on the extracted features. This process uses an algorithm trained on previous datasets to detect speech habits such as quick speech, stammers, and fillers. The input is the speech features, and the output is the identified speaker's habits.

[1148] Step 7:

[1149] The server applies a generative model to remove the identified habits. Specifically, it uses the GPT-3 model and provides a prompt, such as "Please remove fillers and convert to natural speech." The generative model removes unnecessary fillers and habits and converts the speech to natural speech. The input is speaker habit information and audio data, and the output is natural speech data with the habits removed.

[1150] Step 8:

[1151] The server recreates natural speech data with the quirks removed. In this step, the "Vosk" library is used to convert text data into acoustic data and generate natural speech. The input is the identified audio-text data, and the output is the recreated continuous audio data.

[1152] Step 9:

[1153] The server sends the regenerated audio data to the device, where it is repacketized and transmitted securely via a protocol such as HTTPS. The input is the regenerated continuous audio data, and the output is packetized digital audio data.

[1154] Step 10:

[1155] The terminal receives the modified audio data and plays it back to the user. Specifically, it reconstructs the received packetized data and plays it back to the user through a speaker or earphone. The input is packetized digital audio data, and the output is reproduced audio data that the user can hear.

[1156] (Application example 1)

[1157] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1158] In a typical store, if a store clerk has a habit of speaking in a certain way when interacting with customers, it can cause discomfort to the customer. In particular, speaking too quickly, stammering, and using too many fillers can be obstacles to smooth communication with customers. Unless this issue is resolved, there is a risk that the quality of store service will decline and customer satisfaction will decrease. Therefore, the present invention aims to provide a system that automatically removes the habit of store clerks' speaking in a certain way and converts it into natural and fluent speech, thereby facilitating communication between store clerks and customers and improving customer satisfaction.

[1159] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1160] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, and means for analyzing and correcting the voice of the store clerk to smoothly serve customers in the store. This enables the store clerk to speak naturally and fluently without worrying about their speaking habits.

[1161] "Audio data" refers to data in which audio information is recorded in digital format.

[1162] "Analyze in real time" means processing data as it is entered.

[1163] "Speaker features" are data that represent specific patterns and attributes contained in the speaker's voice.

[1164] "De-embedding" means removing or modifying unwanted features in a speaker's speech (e.g., tongue twisters, fillers, etc.).

[1165] "Natural speech" refers to fluent and smooth speech that does not sound strange to the listener.

[1166] "Playback" means outputting the processed audio data as sound through speakers or headphones.

[1167] A "system" is a structure that includes multiple components that work in conjunction with each other.

[1168] A "store" is a physical location for selling goods and services.

[1169] "Customer service" means the service and interaction activities that store staff engage in with customers.

[1170] "Modification" means to change the original state and adjust it to a desired form.

[1171] The present invention is a system that analyzes voice data of store clerks in real time, automatically removes speaking habits, and reproduces natural speech in order to facilitate smooth customer service in stores. This system uses a device such as a smartphone or earphones with a microphone to capture the voice of the store clerk, and processes the voice data on a server. Specific embodiments of the present invention are described below.

[1172] System Configuration

[1173] The system consists of the following main components:

[1174] Terminal: A device that allows the store clerk to input voice and transmit the voice data to the server. Specifically, a smartphone or earphones with a microphone are used.

[1175] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and reproduces natural speech.

[1176] Store Clerk: The subject who inputs speech and receives modified natural speech audio.

[1177] Program Processing Overview

[1178] The program of this system executes the following processes.

[1179] 1. The terminal captures the clerk's voice in real time and sends it to the server.

[1180] 2. The server receives the audio data and performs preprocessing such as noise reduction and volume adjustment.

[1181] 3. The server extracts features from the speech data, including features such as rapid speech, stammers, and fillers.

[1182] 4. The server identifies the speaker's habits based on the extracted features.

[1183] 5. The server applies generative AI models to remove quirks, such as removing fillers and converting fast-paced parts to the appropriate speed.

[1184] 6. The server recreates natural-sounding speech based on the data after the habits have been removed.

[1185] 7. The server sends the regenerated audio data to the device.

[1186] 8. The terminal plays the corrected voice data to the store clerk.

[1187] Hardware used

[1188] Smartphone: A device for voice input and playback.

[1189] Earphones with microphone: A device for high-quality audio capture.

[1190] Software used

[1191] Python: The base programming language.

[1192] SpeechRecognition: A library for speech capture and recognition.

[1193] requests: An HTTP client library for sending audio data to a server.

[1194] Specific examples

[1195] For example, if a store clerk says, "Um, well, I'd like to have a moment of your time," the device captures this speech and sends it to the server in real time. The server analyzes the speech data, identifies fillers such as "um" and "I see," and removes them to convert it into natural speech, "I'd like to have a moment of your time." The server then sends the regenerated speech data to the device, which plays it back to the store clerk. This allows the store clerk to interact with the customer as if they were speaking fluently.

[1196] Prompt Sentence Examples

[1197] For the generative AI model that "removes fillers" and "converts to natural speech," use prompts like these:

[1198] It analyzes the voice data, identifies and removes unnecessary fillers (such as "um" or "hmm"), converts the speech into natural-sounding speech, and converts fast-paced speech to an appropriate speed, outputting smooth speech.

[1199] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1200] Step 1:

[1201] The user inputs voice using earphones with a microphone or the microphone on a smartphone. At this time, the user conducts normal conversation and the voice data is captured. The input voice is stored as raw data on the device.

[1202] Step 2:

[1203] The device transmits the captured audio data to the server in real time. Specifically, the device converts the audio data into a digital format and sends the data to the server using an HTTP request, with the audio data remaining raw.

[1204] Step 3:

[1205] The server receives the voice data sent from the device, converts the received data into an analyzable format, and uses it in the next processing step.

[1206] Step 4:

[1207] The server preprocesses the audio data. Specifically, it performs noise reduction and volume adjustment. This processing improves the quality of the audio data, making it easier to perform subsequent analysis. The input is raw audio data, and the output is audio data with noise reduced and volume adjusted.

[1208] Step 5:

[1209] The server extracts speaker features from the audio data. For example, it uses a speech recognition library to identify features such as rapid speech, stammers, and fillers from the audio signal. At this stage, features are generated as numerical data or in the form of tags.

[1210] Step 6:

[1211] The server identifies the speaker's habits based on the extracted features. For example, if there are many fillers, it will tag it as a "filler habit," and if the speaker speaks quickly, it will identify it as a "fast speaking habit." In this step, the habit data is output as the result of the feature analysis.

[1212] Step 7:

[1213] The server applies a generative AI model to remove the identified habits. Specifically, it modifies the voice data using prompts for the generative AI model ("Analyze the voice data, identify and remove unnecessary fillers (such as 'um' or 'hmm'), and convert it into natural speech. Convert fast-paced parts to an appropriate speed and output smooth speech."). The input is the voice data with identified habits, and the output is the modified natural speech data.

[1214] Step 8:

[1215] The server sends the voice data regenerated into natural speech to the device. The corrected voice data is sent back to the device using an HTTP request. The input is the natural speech data, and the output is the completion status of transmission to the device.

[1216] Step 9:

[1217] The terminal plays the corrected voice data to the user. The clerk can hear fluent and natural speech through a playback device (earphones or smartphone speakers). The input is the corrected voice data received from the server, and the output is the voice that the user can hear.

[1218] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1219] The present invention relates to a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. The system further includes an emotion engine that recognizes the user's emotions and adjusts the method of modifying the voice data based on the emotions.

[1220] System Configuration

[1221] The system consists of the following main components:

[1222] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[1223] Server: A processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech.

[1224] User: The entity that inputs speech and receives modified, natural-speech audio.

[1225] Emotion engine: A module that recognizes the user's emotions in real time and feeds the analysis results back to the system.

[1226] Program processing

[1227] The specific processing of the system will be explained below.

[1228] Input and transmission of voice data

[1229] The user inputs voice using earphones with a microphone or a PC microphone. This voice is normal conversation.

[1230] The terminal captures the audio data in real time and stores the data in a buffer.

[1231] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1232] The device sends the audio data to the server in real time whenever possible.

[1233] Voice data analysis and emotion recognition

[1234] The server receives the voice data transmitted from the terminal.

[1235] The server performs pre-processing of the received audio data, including noise reduction and volume adjustment.

[1236] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1237] The server analyzes the extracted features and identifies speaking habits.

[1238] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[1239] Habit removal and emotional adjustment

[1240] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[1241] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1242] Reproduction and playback of spontaneous speech

[1243] The server recreates natural speech based on the modified voice data.

[1244] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1245] The server transmits the regenerated voice data to the terminal in real time.

[1246] The terminal receives the modified natural speech audio and plays it back to the user.

[1247] Specific examples

[1248] Call Scenarios

[1249] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1250] The device captures this audio in real time and stores it in a buffer.

[1251] The device sends the captured audio data to the server.

[1252] The server receives the audio data and performs noise reduction and volume adjustment.

[1253] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1254] The server identifies the speech habits based on the identified feature amount.

[1255] The emotion engine recognizes that the user is feeling a little nervous and sends that information to the server.

[1256] The server will remove identified fillers such as "uh" and "right," convert fast-paced parts to an appropriate speed, and adjust the volume slightly to provide a sense of security.

[1257] The server reproduces the natural utterance "Please give me a moment."

[1258] The server sends the regenerated audio to the terminal, which plays it back to the user.

[1259] This invention enables natural and smooth communication even for users who are unsure about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses are realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[1260] The processing flow will be explained below.

[1261] Specific processing steps of the program

[1262] Step 1:

[1263] The user inputs audio using earphones with a microphone or the microphone on the PC.

[1264] The terminal captures the audio data in real time and stores the data in a buffer.

[1265] Step 2:

[1266] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1267] The device sends the audio data to the server in real time whenever possible.

[1268] Step 3:

[1269] The server receives the voice data transmitted from the terminal.

[1270] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[1271] The server adjusts the volume to set the optimal listening level.

[1272] Step 4:

[1273] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1274] The server analyzes the extracted features and identifies speaking habits.

[1275] Step 5:

[1276] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[1277] The emotion engine feeds back the recognized emotion information to the server.

[1278] Step 6:

[1279] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[1280] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1281] Step 7:

[1282] The server reproduces natural-spoken speech with the quirks removed and the emotions reflected.

[1283] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1284] Step 8:

[1285] The server transmits the regenerated voice data to the terminal in real time.

[1286] The terminal receives the modified natural speech audio and plays it back to the user.

[1287] Specific examples

[1288] Call Scenarios

[1289] Step 1:

[1290] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1291] The device captures this audio in real time and stores it in a buffer.

[1292] Step 2:

[1293] The device sends the captured audio data to the server.

[1294] Step 3:

[1295] The server receives the audio data and performs noise reduction and volume adjustment.

[1296] Step 4:

[1297] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1298] The server identifies the speech habits based on the identified feature amount.

[1299] Step 5:

[1300] The emotion engine recognizes that the user is feeling a little nervous.

[1301] The emotion engine sends the recognized emotion information to the server.

[1302] Step 6:

[1303] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[1304] Based on feedback from the emotion engine, the server makes adjustments such as slowing down the speaking rate and lowering the volume.

[1305] Step 7:

[1306] The server reproduces the natural utterance "Please give me a moment."

[1307] The server converts the reproduced audio data into a clear, audible format.

[1308] Step 8:

[1309] The server transmits the regenerated voice data to the terminal.

[1310] The terminal plays the modified natural speech to the user.

[1311] Example 2

[1312] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1313] Conventional speech recognition systems have had the problem of being unable to completely remove unnatural speech caused by the speaker's unique habits and emotions. Furthermore, they were unable to recognize the speaker's emotions and adjust their speech accordingly, making it difficult to achieve more natural and personalized speech communication. This has made it difficult for users who are unsure about their speaking style or who stutter to communicate smoothly.

[1314] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1315] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time, means for extracting features from the analyzed voice data, means for identifying the speaker's habits based on the extracted features and removing the habits, means for recognizing the user's emotions in real time and sending the information to the analysis means, means for regenerating the voice data from which the habits have been removed and which has been adjusted based on the emotion information as natural speech, and means for playing back the regenerated voice data to the user. This makes it possible to adjust and optimize the voice data based on the speaker's habits and emotions, thereby achieving natural and smooth communication.

[1316] "Means for inputting voice data" refers to equipment or a method by which a user inputs voice data into the system using a microphone or other input device.

[1317] "Means for real-time analysis" refers to algorithms or software that process and analyze input voice data almost instantaneously.

[1318] "Means for extracting features" refers to a process for extracting specific patterns or attributes (e.g., speaking rate, fillers, stammers, etc.) from speech data.

[1319] "Means for identifying and removing speaker habits" refers to technology that detects a speaker's specific habits (e.g., speaking quickly, using fillers, etc.) based on extracted features and corrects or removes them.

[1320] The "means for recognizing the user's emotions in real time and transmitting that information to the analysis means" is a module for identifying the user's emotional state (e.g., joy, anger, sadness, surprise, etc.) from voice data in real time and reflecting that information in voice analysis.

[1321] A "regenerating means" is an algorithm or device that uses the modified speech data to reconstruct natural speech.

[1322] The "playback means" refers to an audio output device such as a speaker or earphones that allows the user to hear the regenerated audio data.

[1323] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, regenerates natural speech, and further recognizes the user's emotions and corrects the voice data based on those emotions. An embodiment of the present invention consists of the following main components and processing flow.

[1324] System Configuration

[1325] 1. Terminal

[1326] A device that allows a user to input voice and transmit the voice data to a server. Specific examples include earphones with a microphone and a microphone connected to a PC.

[1327] 2. Server

[1328] It is a processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech. The server processes the voice data using an advanced generative AI model.

[1329] 3. Emotion Engine

[1330] This module recognizes the user's emotions in real time and feeds the analysis results back to the system. The emotion engine extracts emotions such as joy, anger, sadness, and surprise from voice data.

[1331] 4. Users

[1332] A subject that inputs speech and receives modified, natural-speech audio.

[1333] Program processing

[1334] Input and transmission of voice data

[1335] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is captured in real time by the device and stored in a buffer. The device then prepares the voice data for transmission and starts sending it to the server. Transmission is performed as quickly as possible in real time.

[1336] Receiving and preprocessing audio data

[1337] The server receives the audio data sent from the device, applies a noise reduction filter to the received audio data to remove background noise, and performs volume equalization to adjust and maintain a constant volume level.

[1338] Voice data analysis and emotion recognition

[1339] The server extracts features from the preprocessed voice data. These features include speaking rate, fillers (e.g., "um," "ah," etc.), and stammers. The server analyzes the features of the voice data and identifies speaking habits. At the same time, the emotion engine recognizes the user's emotions from the voice data in real time and sends that information to the server.

[1340] Habit removal and emotional adjustment

[1341] The server generates a dataset for removing habits based on the identified habits and emotional information. Specifically, it uses a generative AI model to modify the speech data, removing identified fillers and adjusting fast-paced parts to an appropriate speed. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1342] Reproduction and playback of spontaneous speech

[1343] The server uses the modified voice data to recreate natural-sounding speech. The recreated voice data is passed through a sound quality optimization engine, which converts it into a clear, easy-to-listen format. Finally, the server transmits the optimized voice data to the device in real time, and the device plays the received voice back to the user.

[1344] Specific examples

[1345] Call Scenarios

[1346] During a call, a user utters, "Um, well, I'd like to have a moment of your time." The device captures this audio in real time and stores it in a buffer. The device then sends the stored audio data to the server. The server receives the audio data and performs noise reduction and volume adjustment. Features such as speech rate, fillers, and stammers are extracted from the audio data and speech habits are identified based on these. At the same time, the emotion engine recognizes whether the user is feeling nervous and sends this information to the server. The server then removes the identified fillers, adjusts the fast-talking parts to an appropriate speed, and adjusts the volume to provide a sense of security. The regenerated, natural utterance, "I'd like to have a moment of your time," is then sent to the device, which plays it back to the user.

[1347] This system enables natural and smooth communication even for users who are anxious about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses can be realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[1348] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1349] Step 1: Input audio data

[1350] The user inputs voice using earphones with a microphone or the microphone on the PC. At this stage, the user can have a normal conversation.

[1351] Input: User speaking

[1352] Output: Analog audio signal

[1353] Step 2: Capturing and buffering audio data

[1354] The terminal captures the user's input voice in real time and converts it into a digital voice signal.

[1355] The device temporarily stores captured digital audio in a buffer.

[1356] Input: Analog audio signal

[1357] Output: Digital audio data

[1358] Step 3: Prepare to send audio data

[1359] The terminal reads the voice data stored in the buffer at regular intervals and prepares it for transmission, performing preprocessing such as data compression.

[1360] Input: Buffered digital audio data

[1361] Output: Audio data format that can be sent to the server

[1362] Step 4: Sending audio data

[1363] The data is converted into a voice data format that can be transmitted by the terminal and sent to the server. Transmission is performed as quickly as possible in real time.

[1364] Input: Audio data format that can be sent to the server

[1365] Output: Audio data sent to the server

[1366] Step 5: Receiving audio data

[1367] The server receives the voice data transmitted from the terminal in real time.

[1368] Input: Audio data sent from the device

[1369] Output: Received audio data stored on the server

[1370] Step 6: Preprocessing the audio data

[1371] The server applies a noise reduction filter to remove background noise from the audio data.

[1372] The server performs volume equalization to keep the audio volume level constant.

[1373] Input: Received audio data stored on the server

[1374] Output: Audio data with noise removed and volume leveled

[1375] Step 7: Extracting features from audio data

[1376] The server extracts features from the preprocessed speech data, including speaking rate, filler words, and stammers.

[1377] Input: Audio data with noise removed and volume leveled

[1378] Output: Audio data with extracted features

[1379] Step 8: Identify the speaker's habits

[1380] The server identifies speaker habits based on the extracted features, including rapid speech, frequent use of filler words, and stammering.

[1381] Input: Audio data with extracted features

[1382] Output: Data with speaker habits identified

[1383] Step 9: Emotion Recognition

[1384] The emotion engine recognizes the user's emotions in real time from the voice data, including joy, anger, sadness, surprise, etc.

[1385] Input: Audio data with extracted features and identified habits

[1386] Output: User's emotional information

[1387] Step 10: Emotional feedback

[1388] The emotion engine transmits the recognized emotion information to the server.

[1389] Input: User's emotional information

[1390] Output: Emotion information sent to the server

[1391] Step 11: Habit Elimination and Emotional Adjustments

[1392] Based on the identified habits and emotional information, the server generates a dataset for habit removal, which involves applying a generative AI model.

[1393] The server adjusts the speed of fast-paced speech, removes fillers, and adjusts speech speed and volume based on emotional information.

[1394] Input: Data with speaker habits identified and emotional feedback

[1395] Output: Corrected and adjusted audio data

[1396] Step 12: Regenerating spontaneous speech

[1397] The server recreates natural speech based on the corrected and adjusted voice data.

[1398] The server runs the regenerated audio data through a sound quality optimization engine to convert it into a clear, audible format.

[1399] Input: Corrected and adjusted audio data

[1400] Output: Regenerated and optimized audio data

[1401] Step 13: Sending the Regenerated Audio

[1402] The server transmits the regenerated voice data to the terminal in real time.

[1403] Input: Regenerated and optimized audio data

[1404] Output: Audio data sent to the device

[1405] Step 14: Playing back audio data

[1406] The terminal receives the regenerated voice data and plays it back to the user, so that the user hears modified, natural-sounding speech.

[1407] Input: Audio data sent to the device

[1408] Output: The audio played to the user

[1409] (Application example 2)

[1410] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1411] When interacting with customers in real-world situations, customer support staff are required to respond immediately to their customers' emotions and reactions, but they lack the technology to appropriately recognize the speaker's habits and emotions and adjust their speech in real time. This can make natural communication difficult, which can lead to a decrease in customer satisfaction.

[1412] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1413] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, means for recognizing emotions and adjusting a method for correcting the voice data based on the recognition results, and means for performing voice analysis and emotion recognition when interacting with a customer in real space and supporting a response according to the customer's emotions. This allows voice to be regenerated as natural speech in real time in accordance with the customer's emotions, enabling customer support staff to communicate more smoothly.

[1414] "Sound data" refers to an acoustic signal expressed in audio format.

[1415] "Input means" refers to a hardware or software mechanism for inputting voice data into the system.

[1416] The "analyzing means" is a component that has the function of processing input voice data and extracting specific features.

[1417] "Features" are data extracted from speech data that indicate the speaker's habits and speech characteristics.

[1418] The "means for identifying habits" is a device that has the function of identifying specific phrases, fillers, etc. from a speaker's speech.

[1419] The "means for removing" is a system that has the function of performing processing to remove the habits of the identified speaker.

[1420] The "regenerating means" is a mechanism for generating natural speech based on speech data from which habits have been removed.

[1421] A "means for playing" is a device for providing a user with reproduced natural speech sounds.

[1422] The "means for recognizing emotions" is a component that has the function of identifying emotions from the user's voice data and feeding that information back to the system.

[1423] The "means for adjusting the correction method" is a mechanism having a function for dynamically changing the method for correcting voice data based on the emotion recognition result.

[1424] "Real space" refers to the physical world, the environment in which a user actually exists and acts.

[1425] A "customer" is an entity that uses a service and is the person with whom the service is interacted.

[1426] "Voice analysis during dialogue" is the process of analyzing voice data in real time during a conversation with a customer and understanding its content.

[1427] The "means for emotion recognition" is a mechanism for analyzing and recognizing customer emotions in real time from voice data during a conversation.

[1428] The "means for supporting the response" is a device that has the function of supporting the customer support staff in order to respond appropriately according to the customer's feelings.

[1429] This invention is applied to a customer support system that handles customer support in the real world, and analyzes voice data in real time, removing the speaker's habits and regenerating natural speech. In addition, by using an emotion engine, the content of speech is adjusted based on the customer's emotions.

[1430] System Configuration

[1431] The system consists of the following main components:

[1432] Terminal: A device that inputs voice data and transmits the data to a server. As a concrete example, smart glasses are used.

[1433] Server: Analyzes voice data, extracts features, identifies and removes habits, reproduces natural speech, and recognizes emotions. This uses the Google Speech Recognition API and Transformers library.

[1434] User: The entity that inputs voice data through the system and receives modified, natural-sounding speech. Specifically, customer support staff.

[1435] Hardware configuration and data calculation

[1436] 1. Capture and transmit audio data

[1437] The device captures audio data through a microphone built into the smart glasses, which is temporarily buffered within the device and then transmitted to the server in real time.

[1438] 2. Voice Recognition

[1439] The server converts the voice data received from the device into text using the Google Speech Recognition API, which provides highly accurate voice recognition.

[1440] 3. Emotion recognition

[1441] The server uses an emotion engine (a model using the Transformers library) based on the converted text to identify the customer's emotions, which typically include emotion categories such as joy, anger, and sadness.

[1442] 4. Correcting speech habits

[1443] The server processes the text to identify and remove slurred speech, fillers, rapid speech, and other habits. The removed features are saved as the corrected text.

[1444] 5. Reproducing natural speech

[1445] The server then regenerates the corrected text and produces natural-sounding speech, adjusting the volume and speech rate based on the emotional information obtained from the emotion engine.

[1446] 6. Playing the corrected audio

[1447] The terminal receives the modified voice data sent from the server and plays it back to the user (customer support staff) through the smart glasses' speaker.

[1448] Specific examples

[1449] Physical store scenario

[1450] 1. Customer: "Um, can you tell me what features this product has?"

[1451] 2. The smart glasses capture the customer's speech and send the voice data to a PC.

[1452] 3. The PC receives the voice data and converts it into text using the Google Speech Recognition API.

[1453] 4. The server performs sentiment analysis on the converted text and recognizes the sentiment of "high interest."

[1454] 5. The program corrected the sentence by removing the "um" from "Um, can you tell me what features this product has?"

[1455] 6. Revised text: "Can you tell me what features this product has?"

[1456] 7. The regenerated audio is played back to the customer support staff through the smart glasses' speakers.

[1457] Prompt Sentence Examples

[1458] Question: "Please explain how the program can correct and regenerate speech when a customer says, 'Um, can you tell me what features this product has?'"

[1459] Example input: Customer: "Um, can you tell me what features this product has?"

[1460] Example output: Regenerated speech: "Can you tell me what features this product has?"

[1461] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1462] Step 1:

[1463] Capture and transmit audio data

[1464] A microphone built into the user's smart glasses captures the customer's speech in real time, generating digital audio data. This audio data is temporarily buffered inside the device and then sent to the server via the device's communication function. The input is the customer's speech, and the output is digital audio data sent to the server.

[1465] Step 2:

[1466] Voice Recognition

[1467] The server processes the voice data received from the device and converts it into text using the Google Speech Recognition API. This process results in the content of the voice data being expressed in text form. The input is the voice data sent to the server, and the output is the corresponding text data.

[1468] Step 3:

[1469] emotion recognition

[1470] The server analyzes emotions based on the converted text data using the Transformers library. It uses a generative AI model to analyze emotions from the customer's speech and obtains the results. The input is the converted text data, and the output is the customer's emotional data (e.g., joy, anger, sadness, etc.).

[1471] Step 4:

[1472] Correcting speech habits

[1473] The server identifies and removes speaker habits (fillers, slurred speech, rapid speech, etc.) from the text data along with the emotion recognition results. At this stage, unnecessary fillers and meaningless words are removed from the text. The inputs are the emotion recognition results and the text data, and the output is the corrected text data.

[1474] Step 5:

[1475] Natural speech reproduction

[1476] The server regenerates natural-sounding speech based on the corrected text data and emotion recognition results, adjusting the volume and speech rate. During this process, the text data is converted into clear speech data. The inputs are the corrected text data and emotion recognition results, and the output is the regenerated speech data.

[1477] Step 6:

[1478] Playing the corrected audio

[1479] The terminal receives the regenerated voice data sent from the server and plays it back to the user through the smart glasses speaker. In this process, the customer support staff as the user can hear the corrected natural-sounding speech. The input is the regenerated voice data, and the output is the reproduced voice provided to the user.

[1480] At each step, appropriate data processing or calculation is performed based on specific input data, and the output data required for the next step is generated as a result, thereby realizing natural communication with customers.

[1481] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1482] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1483] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1484] [Fourth embodiment]

[1485] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1486] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1487] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1488] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1489] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1490] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1491] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1492] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1493] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1494] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1495] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1496] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1497] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1498] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used with devices such as telephones, web conferences, and earphones with microphones. Specific embodiments of the present invention are described below.

[1499] System Configuration

[1500] The system consists of the following main components:

[1501] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[1502] Server: A processing device that receives voice data, analyzes it in real time, removes habits, and reproduces natural speech.

[1503] User: The entity that inputs speech and receives modified, natural-speech audio.

[1504] Program processing

[1505] The operation of the system will be specifically described below.

[1506] Input and transmission of voice data

[1507] 1. The user inputs audio using earphones with a microphone or a PC microphone. This audio is normal conversation.

[1508] 2. The device captures the audio data in real time and prepares to send it to the server.

[1509] 3. The device sends the captured audio data to the server.

[1510] Analyzing voice data and identifying habits

[1511] 4. The server receives the voice data sent from the device.

[1512] 5. The server performs preprocessing on the received audio data, including noise reduction and volume adjustment.

[1513] 6. The server extracts features from the audio data, including features such as rapid speech, stammers, and fillers.

[1514] 7. The server identifies the speaker's habits based on the extracted features.

[1515] Removes quirks and reproduces natural speech

[1516] 8. The server applies generative AI models to remove identified habits, e.g., remove fillers and convert fast-paced parts to an appropriate speed.

[1517] 9. The server recreates natural-sounding speech based on the data after the habits have been removed.

[1518] Sending and playing back modified audio

[1519] 10. The server sends the regenerated audio data to the device.

[1520] 11. The device plays the modified audio data to the user, allowing the user to hear natural, fluent speech.

[1521] Specific examples

[1522] A specific example of use is shown below.

[1523] Call Scenarios

[1524] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1525] The device captures this audio and transmits it to the server in real time.

[1526] The server analyzes the audio data and identifies fillers such as "um" and "right."

[1527] The server removes the filler and converts it into natural speech, such as "I'd like to ask for a moment of your time."

[1528] The server sends the regenerated audio to the terminal, which plays it back.

[1529] The user can continue the conversation as if they were speaking fluently while listening to the other party's responses.

[1530] This invention enables users who are unsure about their speaking style or who stutter to communicate naturally and smoothly. This system will be an extremely useful tool in both business and personal settings.

[1531] The processing flow will be explained below.

[1532] Specific processing steps of the program

[1533] Step 1:

[1534] The user inputs audio using earphones with a microphone or the microphone on the PC.

[1535] The terminal captures the audio data in real time and stores the data in a buffer.

[1536] Step 2:

[1537] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1538] The device sends the audio data to the server in real time whenever possible.

[1539] Step 3:

[1540] The server receives the voice data transmitted from the terminal.

[1541] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[1542] Step 4:

[1543] The server adjusts the volume to set the optimal listening level.

[1544] The server completes preprocessing in order to analyze the audio data in real time.

[1545] Step 5:

[1546] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1547] The server analyzes the extracted features and identifies speaking habits.

[1548] Step 6:

[1549] The server generates a dataset to correct the identified habits.

[1550] The server applies a generative AI model to make corrections such as adjusting fast-talking parts and removing filler.

[1551] Step 7:

[1552] The server recreates natural speech based on the modified voice data.

[1553] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1554] Step 8:

[1555] The server transmits the regenerated voice data to the terminal in real time.

[1556] The terminal receives the modified natural speech audio and plays it back to the user.

[1557] Specific examples

[1558] Call Scenarios

[1559] Step 1:

[1560] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1561] The device captures this audio in real time and stores it in a buffer.

[1562] Step 2:

[1563] The device sends the captured audio data to the server.

[1564] Step 3:

[1565] The server receives the audio data and performs noise reduction.

[1566] Step 4:

[1567] The server adjusts the volume.

[1568] Step 5:

[1569] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1570] The server identifies the speech habits based on the identified feature amount.

[1571] Step 6:

[1572] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[1573] Step 7:

[1574] The server reproduces the natural utterance "Please give me a moment."

[1575] Step 8:

[1576] The server sends the regenerated audio to the terminal, which plays it back to the user.

[1577] Example 1

[1578] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1579] Conventional speech data analysis systems have had difficulty accurately removing speaker habits and regenerating natural speech. In particular, there is a need to provide natural conversations by removing characteristic habits such as rapid speech and fillers in real time. To solve this problem, an efficient and accurate method for analyzing and regenerating speech data is required.

[1580] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1581] In this invention, the server includes means for receiving speech data and performing preprocessing such as noise reduction and volume adjustment, means for extracting features from the speech data and identifying the speaker's habits, and means for applying a generative model to remove the identified habits and regenerate natural speech. This makes it possible to generate natural speech in real time while removing the speaker's unnatural habits.

[1582] "Voice data" refers to voice information stored in digital format that a user makes through an input device such as a microphone.

[1583] "Real-time" refers to data and events being processed as they occur, with immediate results available without delay.

[1584] "Noise reduction" is a technology for improving the clarity of audio by removing unnecessary background sounds and noise from audio data.

[1585] "Volume adjustment" is a process for equalizing the volume level of recorded audio data to ensure that it is easy to listen to when played back.

[1586] "Features" refer to specific patterns or attributes extracted from speech data, and are information used to identify a speaker's habits and characteristics.

[1587] A "generative model" is an artificial intelligence model used to create new data from existing data, and in this case is specifically used to regenerate audio data.

[1588] "Fillers" refer to unnecessary parts of speech, such as "ums" and "hmms," that speakers unconsciously use to connect words.

[1589] "Habits" refer to a speaker's particular speech patterns or habitual speaking characteristics that need to be identified and eliminated.

[1590] "Regeneration" refers to the process of reconstructing processed data in its original or new form, which in this case means producing natural-sounding speech.

[1591] This system analyzes input speech data in real time, automatically removes the speaker's habits, and reproduces natural speech. This system can be used on devices such as telephones, web conferences, and earphones with microphones.

[1592] System Configuration

[1593] The system consists of the following main components:

[1594] Terminal: A device that allows a user to input voice and transmit the voice data to a server. Examples include smartphones, PCs, and tablets.

[1595] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and regenerates natural speech. The server is equipped with a high-performance processor and sufficient memory.

[1596] User: The entity that inputs speech and receives modified, natural-speech audio.

[1597] Hardware and Software

[1598] The following hardware and software are used to implement this system.

[1599] Hardware:

[1600] Earphones with a microphone or a microphone built into your PC: Captures your voice.

[1601] Devices such as smartphones, PCs, and tablets: Captures audio data in real time and sends it to a server.

[1602] Server equipped with a high-performance processor: Analyzes and processes voice data.

[1603] software:

[1604] Audio processing library "SoX": Performs preprocessing for noise reduction and volume adjustment.

[1605] Machine learning library "TensorFlow": Extracts features from audio data and identifies habits.

[1606] Generative AI model "GPT-3": Removes fillers and reproduces natural speech.

[1607] "Vosk" speech generation library: converts text data into acoustic data.

[1608] Specific examples

[1609] Below are some specific examples of how the system can be used.

[1610] Call Scenarios

[1611] 1. During a call, the user says, "Um, I'd like to have a moment of your time."

[1612] 2. The device captures this audio and sends it to the server in real time.

[1613] 3. The server analyzes the audio data and identifies fillers such as "uh" and "right."

[1614] 4. The server removes the filler and converts the utterance into natural speech: "Please give me a moment of your time."

[1615] 5. The server sends the regenerated audio to the device, which plays it back.

[1616] 6. Users can continue the conversation as if they were speaking fluently while listening to the responses of the other party.

[1617] Prompt Sentence Examples

[1618] Below are some examples of prompts that can be given to the generative AI model.

[1619] Prompt Sentence Examples

[1620] Input: "Um, I'd like to, well, take a moment."

[1621] Instructions to the generative AI model: "Remove fillers and convert to natural-sounding speech."

[1622] In this way, a system is provided that allows the speaker to have a smooth conversation.

[1623] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1624] Step 1:

[1625] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is a normal conversation, such as "Um, I'd like to have a moment of your time." The input voice data is captured as an analog signal and converted to a digital signal within the device.

[1626] Step 2:

[1627] The device stores the captured audio data in memory in real time and prepares it for transmission to the server. During this process, the audio data is divided into packets at short time intervals (for example, every second). The data is then sent to the server using a secure communication protocol such as HTTPS. The input is analog audio data, and the output is packetized digital audio data.

[1628] Step 3:

[1629] The server receives the audio data sent from the terminal. The received packets are reassembled to form a continuous audio data stream. The input is packetized digital audio data, and the output is a reassembled continuous audio data stream.

[1630] Step 4:

[1631] The server performs pre-processing for noise reduction and volume adjustment on the received audio data. This process uses the "SoX" library to remove background noise and equalize the volume. Specifically, it applies noise filtering and audio normalization algorithms. The input is a reconstructed continuous audio data stream, and the output is pre-processed, clean audio data.

[1632] Step 5:

[1633] The server extracts features from the preprocessed audio data. In this step, the machine learning library "TensorFlow" is used to calculate features such as Mel-Frequency Cepstrum Coefficients (MFCCs) from the audio data. The features represent specific attributes of the audio data and are used as input data to identify the speaker's habits. The input is preprocessed, clean audio data, and the output is the extracted audio features.

[1634] Step 6:

[1635] The server identifies the speaker's habits based on the extracted features. This process uses an algorithm trained on previous datasets to detect speech habits such as quick speech, stammers, and fillers. The input is the speech features, and the output is the identified speaker's habits.

[1636] Step 7:

[1637] The server applies a generative model to remove the identified habits. Specifically, it uses the GPT-3 model and provides a prompt, such as "Please remove fillers and convert to natural speech." The generative model removes unnecessary fillers and habits and converts the speech to natural speech. The input is speaker habit information and audio data, and the output is natural speech data with the habits removed.

[1638] Step 8:

[1639] The server recreates natural speech data with the quirks removed. In this step, the "Vosk" library is used to convert text data into acoustic data and generate natural speech. The input is the identified audio-text data, and the output is the recreated continuous audio data.

[1640] Step 9:

[1641] The server sends the regenerated audio data to the device, where it is repacketized and transmitted securely via a protocol such as HTTPS. The input is the regenerated continuous audio data, and the output is packetized digital audio data.

[1642] Step 10:

[1643] The terminal receives the modified audio data and plays it back to the user. Specifically, it reconstructs the received packetized data and plays it back to the user through a speaker or earphone. The input is packetized digital audio data, and the output is played-back audio that the user can hear.

[1644] (Application example 1)

[1645] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1646] In a typical store, if a store clerk has a habit of speaking in a certain way when interacting with customers, it can cause discomfort to the customer. In particular, speaking too quickly, stammering, and using too many fillers can be obstacles to smooth communication with customers. Unless this issue is resolved, there is a risk that the quality of store service will decline and customer satisfaction will decrease. Therefore, the present invention aims to provide a system that automatically removes store clerks' habitual speech patterns and converts them into natural, fluent speech, thereby facilitating communication between store clerks and customers and improving customer satisfaction.

[1647] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1648] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, and means for analyzing and correcting the voice of the store clerk to smoothly serve customers in the store. This enables the store clerk to speak naturally and fluently without worrying about their speaking habits.

[1649] "Audio data" refers to data in which audio information is recorded in digital format.

[1650] "Analyze in real time" means processing data as it is entered.

[1651] "Speaker features" are data that represent specific patterns and attributes contained in the speaker's voice.

[1652] "De-embedding" means removing or modifying unwanted features in a speaker's speech (e.g., tongue twisters, fillers, etc.).

[1653] "Natural speech" refers to fluent and smooth speech that does not sound strange to the listener.

[1654] "Playback" means outputting the processed audio data as sound through speakers or headphones.

[1655] A "system" is a structure that includes multiple components that work in conjunction with each other.

[1656] A "store" is a physical location for selling goods and services.

[1657] "Customer service" refers to the service and interaction activities that store staff engage in with customers.

[1658] "Modification" means to change the original state and adjust it to a desired form.

[1659] The present invention is a system that analyzes voice data of store clerks in real time, automatically removes speaking habits, and reproduces natural speech in order to facilitate smooth customer service in stores. This system captures the voice of the store clerk using a device such as a smartphone or earphones with a microphone, and processes the voice data on a server. Specific embodiments of the present invention are described below.

[1660] System Configuration

[1661] The system consists of the following main components:

[1662] Terminal: A device that allows the store clerk to input voice and transmit the voice data to the server. Specifically, a smartphone or earphones with a microphone are used.

[1663] Server: A processing device that receives voice data, analyzes it in real time, removes quirks, and reproduces natural speech.

[1664] Store Clerk: The subject who inputs speech and receives modified natural speech audio.

[1665] Program Processing Overview

[1666] The program of this system executes the following processes.

[1667] 1. The terminal captures the clerk's voice in real time and sends it to the server.

[1668] 2. The server receives the audio data and performs preprocessing such as noise reduction and volume adjustment.

[1669] 3. The server extracts features from the speech data, including features such as rapid speech, stammers, and fillers.

[1670] 4. The server identifies the speaker's habits based on the extracted features.

[1671] 5. The server applies generative AI models to remove quirks, e.g., remove fillers and convert fast-paced parts to the appropriate speed.

[1672] 6. The server recreates natural-sounding speech based on the data after the habits have been removed.

[1673] 7. The server sends the regenerated audio data to the device.

[1674] 8. The terminal plays the corrected voice data to the store clerk.

[1675] Hardware used

[1676] Smartphone: A device for voice input and playback.

[1677] Earphones with microphone: A device for high-quality audio capture.

[1678] Software used

[1679] Python: The base programming language.

[1680] SpeechRecognition: A library for speech capture and recognition.

[1681] requests: An HTTP client library for sending audio data to a server.

[1682] Specific examples

[1683] For example, if a store clerk says, "Um, well, I'd like to have a moment of your time," the device captures this speech and sends it to the server in real time. The server analyzes the speech data, identifies fillers such as "um" and "I see," and removes them to convert it into natural speech, "I'd like to have a moment of your time." The server then sends the regenerated speech data to the device, which plays it back to the store clerk. This allows the store clerk to interact with the customer as if they were speaking fluently.

[1684] Prompt Sentence Examples

[1685] For the generative AI model that "removes fillers" and "converts to natural speech," use prompts like these:

[1686] It analyzes the voice data, identifies and removes unnecessary fillers (such as "um" or "hmm"), converts the speech into natural-sounding speech, and converts fast-paced speech to an appropriate speed, outputting smooth speech.

[1687] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1688] Step 1:

[1689] The user inputs voice using earphones with a microphone or the microphone on a smartphone. At this time, the user conducts normal conversation and the voice data is captured. The input voice is stored as raw data on the device.

[1690] Step 2:

[1691] The device transmits the captured audio data to the server in real time. Specifically, the device converts the audio data into a digital format and sends the data to the server using an HTTP request, with the audio data remaining raw.

[1692] Step 3:

[1693] The server receives the voice data sent from the device, converts the received data into an analyzable format, and uses it in the next processing step.

[1694] Step 4:

[1695] The server preprocesses the audio data. Specifically, it performs noise reduction and volume adjustment. This processing improves the quality of the audio data, making it easier to perform subsequent analysis. The input is raw audio data, and the output is audio data with noise reduced and volume adjusted.

[1696] Step 5:

[1697] The server extracts speaker features from the audio data. For example, it uses a speech recognition library to identify features such as rapid speech, stammers, and fillers from the audio signal. At this stage, features are generated as numerical data or in the form of tags.

[1698] Step 6:

[1699] The server identifies the speaker's habits based on the extracted features. For example, if there are many fillers, it will tag them as a "filler habit," and if they speak quickly, it will identify them as a "fast speaking habit." In this step, the habit data is output as the result of feature analysis.

[1700] Step 7:

[1701] The server applies a generative AI model to remove the identified habits. Specifically, it modifies the voice data using prompts for the generative AI model ("Analyze the voice data, identify and remove unnecessary fillers (such as 'um' or 'hmm'), and convert it into natural speech. Convert fast-paced parts to an appropriate speed and output smooth speech."). The input is the voice data with identified habits, and the output is the modified natural speech data.

[1702] Step 8:

[1703] The server sends the voice data regenerated into natural speech to the device. The modified voice data is sent back to the device using an HTTP request. The input is the natural speech data, and the output is the completion status of transmission to the device.

[1704] Step 9:

[1705] The terminal plays the corrected voice data to the user. The clerk can hear fluent and natural speech through a playback device (earphones or smartphone speakers). The input is the corrected voice data received from the server, and the output is the voice that the user can hear.

[1706] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1707] The present invention relates to a system that analyzes input voice data in real time, automatically removes the speaker's habits, and reproduces natural speech. The system further includes an emotion engine that recognizes the user's emotions and adjusts the method of modifying the voice data based on the emotions.

[1708] System Configuration

[1709] The system consists of the following main components:

[1710] Terminal: A device through which a user inputs voice and transmits the voice data to a server.

[1711] Server: A processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech.

[1712] User: The entity that inputs speech and receives modified, natural-speech audio.

[1713] Emotion engine: A module that recognizes the user's emotions in real time and feeds the analysis results back to the system.

[1714] Program processing

[1715] The specific processing of the system will be explained below.

[1716] Input and transmission of voice data

[1717] The user inputs voice using earphones with a microphone or a PC microphone. This voice is normal conversation.

[1718] The terminal captures the audio data in real time and stores the data in a buffer.

[1719] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1720] The device sends the audio data to the server in real time whenever possible.

[1721] Voice data analysis and emotion recognition

[1722] The server receives the voice data transmitted from the terminal.

[1723] The server performs pre-processing of the received audio data, including noise reduction and volume adjustment.

[1724] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1725] The server analyzes the extracted features and identifies speaking habits.

[1726] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[1727] Habit removal and emotional adjustment

[1728] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[1729] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1730] Reproduction and playback of spontaneous speech

[1731] The server recreates natural speech based on the modified voice data.

[1732] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1733] The server transmits the regenerated voice data to the terminal in real time.

[1734] The terminal receives the modified natural speech audio and plays it back to the user.

[1735] Specific examples

[1736] Call Scenarios

[1737] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1738] The device captures this audio in real time and stores it in a buffer.

[1739] The device sends the captured audio data to the server.

[1740] The server receives the audio data and performs noise reduction and volume adjustment.

[1741] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1742] The server identifies the speech habits based on the identified feature amount.

[1743] The emotion engine recognizes that the user is feeling a little nervous and sends that information to the server.

[1744] The server will remove identified fillers such as "uh" and "right," convert fast-paced parts to an appropriate speed, and adjust the volume slightly to provide a sense of security.

[1745] The server reproduces the natural utterance "Please give me a moment."

[1746] The server sends the regenerated audio to the terminal, which plays it back to the user.

[1747] This invention enables natural and smooth communication even for users who are unsure about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses are realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[1748] The processing flow will be explained below.

[1749] Specific processing steps of the program

[1750] Step 1:

[1751] The user inputs audio using earphones with a microphone or the microphone on the PC.

[1752] The terminal captures the audio data in real time and stores the data in a buffer.

[1753] Step 2:

[1754] The terminal prepares the stored voice data for transmission and starts transmitting it to the server.

[1755] The device sends the audio data to the server in real time whenever possible.

[1756] Step 3:

[1757] The server receives the voice data transmitted from the terminal.

[1758] The server performs noise reduction on the audio data received to improve the clarity of the audio.

[1759] The server adjusts the volume to set the optimal listening level.

[1760] Step 4:

[1761] The server extracts features from the speech data, including speech speed, stammers, fillers, etc.

[1762] The server analyzes the extracted features and identifies speaking habits.

[1763] Step 5:

[1764] The emotion engine recognizes the user's emotions (e.g., joy, anger, sadness, surprise, etc.) from the voice data.

[1765] The emotion engine feeds back the recognized emotion information to the server.

[1766] Step 6:

[1767] The server generates a dataset for removing habits based on the identified habits and recognized emotions.

[1768] The server applies the generative AI model to adjust fast-talking parts, remove fillers, and make other corrections. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1769] Step 7:

[1770] The server reproduces natural-spoken speech with the quirks removed and the emotions reflected.

[1771] The server optimizes the reproduced audio data and converts it into a clear, audible format.

[1772] Step 8:

[1773] The server transmits the regenerated voice data to the terminal in real time.

[1774] The terminal receives the modified natural speech audio and plays it back to the user.

[1775] Specific examples

[1776] Call Scenarios

[1777] Step 1:

[1778] During a call, the user utters, "Um, I'd like to have a moment of your time, please."

[1779] The device captures this audio in real time and stores it in a buffer.

[1780] Step 2:

[1781] The device sends the captured audio data to the server.

[1782] Step 3:

[1783] The server receives the audio data and performs noise reduction and volume adjustment.

[1784] Step 4:

[1785] The server extracts features such as rapid speech, stammers, and fillers from the audio data.

[1786] The server identifies the speech habits based on the identified feature amount.

[1787] Step 5:

[1788] The emotion engine recognizes that the user is feeling a little nervous.

[1789] The emotion engine sends the recognized emotion information to the server.

[1790] Step 6:

[1791] The server removes identified fillers such as "uh," "right," etc., and converts fast-paced parts to the appropriate speed.

[1792] Based on feedback from the emotion engine, the server makes adjustments such as slowing down the speaking rate and lowering the volume.

[1793] Step 7:

[1794] The server reproduces the natural utterance "Please give me a moment."

[1795] The server converts the reproduced audio data into a clear, audible format.

[1796] Step 8:

[1797] The server transmits the regenerated voice data to the terminal.

[1798] The terminal plays the modified natural speech to the user.

[1799] Example 2

[1800] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1801] Conventional speech recognition systems have had the problem of being unable to completely remove unnatural speech caused by the speaker's unique habits and emotions. Furthermore, they were unable to recognize the speaker's emotions and adjust their speech accordingly, making it difficult to achieve more natural and personalized speech communication. This has made it difficult for users who are unsure about their speaking style or who stutter to communicate smoothly.

[1802] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1803] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time, means for extracting features from the analyzed voice data, means for identifying the speaker's habits based on the extracted features and removing the habits, means for recognizing the user's emotions in real time and sending the information to the analysis means, means for regenerating the voice data from which the habits have been removed and which has been adjusted based on the emotion information as natural speech, and means for playing back the regenerated voice data to the user. This makes it possible to adjust and optimize the voice data based on the speaker's habits and emotions, thereby achieving natural and smooth communication.

[1804] "Means for inputting voice data" refers to equipment or a method by which a user inputs voice data into the system using a microphone or other input device.

[1805] "Means for real-time analysis" refers to algorithms or software that process and analyze input voice data almost instantaneously.

[1806] "Feature extraction" refers to a process for extracting specific patterns or attributes (e.g., speaking rate, fillers, stammers, etc.) from speech data.

[1807] "Means for identifying and removing speaker habits" refers to technology that detects a speaker's specific habits (e.g., speaking quickly, using fillers, etc.) based on extracted features and corrects or removes them.

[1808] The "means for recognizing the user's emotions in real time and transmitting that information to the analysis means" is a module for identifying the user's emotional state (e.g., joy, anger, sadness, surprise, etc.) from voice data in real time and reflecting that information in voice analysis.

[1809] A "regenerating means" is an algorithm or device that uses the modified speech data to reconstruct natural speech.

[1810] The "playback means" refers to an audio output device such as a speaker or earphones that allows the user to hear the regenerated audio data.

[1811] The present invention is a system that analyzes input voice data in real time, automatically removes the speaker's habits, regenerates natural speech, and further recognizes the user's emotions and corrects the voice data based on those emotions. An embodiment of the present invention consists of the following main components and processing flow.

[1812] System Configuration

[1813] 1. Terminal

[1814] A device that allows a user to input voice and transmit the voice data to a server. Specific examples include earphones with a microphone and a microphone connected to a PC.

[1815] 2. Server

[1816] It is a processing device that receives voice data, analyzes it in real time, removes habits, recognizes emotions, and reproduces natural speech. The server processes the voice data using an advanced generative AI model.

[1817] 3. Emotion Engine

[1818] This module recognizes the user's emotions in real time and feeds the analysis results back to the system. The emotion engine extracts emotions such as joy, anger, sadness, and surprise from voice data.

[1819] 4. Users

[1820] A subject that inputs speech and receives modified, natural-speech audio.

[1821] Program processing

[1822] Input and transmission of voice data

[1823] The user inputs voice using earphones with a microphone or a PC microphone. The input voice is captured in real time by the device and stored in a buffer. The device then prepares the voice data for transmission and starts sending it to the server. Transmission is performed as quickly as possible in real time.

[1824] Receiving and preprocessing audio data

[1825] The server receives the audio data sent from the device, applies a noise reduction filter to the received audio data to remove background noise, and performs volume equalization to adjust and maintain a constant volume level.

[1826] Voice data analysis and emotion recognition

[1827] The server extracts features from the preprocessed voice data. These features include speaking rate, fillers (e.g., "um," "ah," etc.), and stammers. The server analyzes the features of the voice data and identifies speaking habits. At the same time, the emotion engine recognizes the user's emotions from the voice data in real time and sends that information to the server.

[1828] Habit removal and emotional adjustment

[1829] The server generates a dataset for removing habits based on the identified habits and emotional information. Specifically, it uses a generative AI model to modify the speech data, removing identified fillers and adjusting fast-paced parts to an appropriate speed. It also adjusts speech rate and volume based on feedback from the emotion engine.

[1830] Reproduction and playback of spontaneous speech

[1831] The server uses the modified voice data to recreate natural-sounding speech. The recreated voice data is passed through a sound quality optimization engine, which converts it into a clear, easy-to-listen format. Finally, the server transmits the optimized voice data to the device in real time, and the device plays the received voice back to the user.

[1832] Specific examples

[1833] Call Scenarios

[1834] During a call, a user utters, "Um, well, I'd like to have a moment of your time." The device captures this audio in real time and stores it in a buffer. The device then sends the stored audio data to the server. The server receives the audio data and performs noise reduction and volume adjustment. Features such as speech rate, fillers, and stammers are extracted from the audio data and speech habits are identified based on these. At the same time, the emotion engine recognizes whether the user is feeling nervous and sends this information to the server. The server then removes the identified fillers, adjusts the fast-talking parts to an appropriate speed, and adjusts the volume to provide a sense of security. The regenerated, natural utterance, "I'd like to have a moment of your time," is then sent to the device, which plays it back to the user.

[1835] This system enables natural and smooth communication even for users who are anxious about their speaking style or who stutter. Furthermore, by combining it with an emotion engine, more personalized responses can be realized, providing communication that takes into consideration not only the content of speech but also the user's emotions.

[1836] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1837] Step 1: Input audio data

[1838] The user inputs voice using earphones with a microphone or the microphone on the PC. At this stage, the user can have a normal conversation.

[1839] Input: User speaking

[1840] Output: Analog audio signal

[1841] Step 2: Capturing and buffering audio data

[1842] The terminal captures the user's input voice in real time and converts it into a digital voice signal.

[1843] The device temporarily stores captured digital audio in a buffer.

[1844] Input: Analog audio signal

[1845] Output: Digital audio data

[1846] Step 3: Prepare to send audio data

[1847] The terminal reads the voice data stored in the buffer at regular intervals and prepares it for transmission, during which preprocessing such as data compression is performed.

[1848] Input: Buffered digital audio data

[1849] Output: Audio data format that can be sent to the server

[1850] Step 4: Sending audio data

[1851] The data is converted into a voice data format that can be transmitted by the terminal and sent to the server. Transmission is performed as quickly as possible in real time.

[1852] Input: Audio data format that can be sent to the server

[1853] Output: Audio data sent to the server

[1854] Step 5: Receiving audio data

[1855] The server receives the voice data transmitted from the terminal in real time.

[1856] Input: Audio data sent from the device

[1857] Output: Received audio data stored on the server

[1858] Step 6: Preprocessing the audio data

[1859] The server applies a noise reduction filter to remove background noise from the audio data.

[1860] The server performs volume equalization to keep the audio volume level constant.

[1861] Input: Received audio data stored on the server

[1862] Output: Audio data with noise removed and volume leveled

[1863] Step 7: Extracting features from audio data

[1864] The server extracts features from the preprocessed speech data, including speaking rate, filler words, and stammers.

[1865] Input: Audio data with noise removed and volume leveled

[1866] Output: Audio data with extracted features

[1867] Step 8: Identify the speaker's habits

[1868] The server identifies speaker habits based on the extracted features, including rapid speech, frequent use of filler words, and stammering.

[1869] Input: Audio data with extracted features

[1870] Output: Data with speaker habits identified

[1871] Step 9: Emotion Recognition

[1872] The emotion engine recognizes the user's emotions in real time from the voice data, including joy, anger, sadness, surprise, etc.

[1873] Input: Audio data with extracted features and identified habits

[1874] Output: User's emotional information

[1875] Step 10: Emotional feedback

[1876] The emotion engine transmits the recognized emotion information to the server.

[1877] Input: User's emotional information

[1878] Output: Emotion information sent to the server

[1879] Step 11: Habit Elimination and Emotional Adjustments

[1880] Based on the identified habits and emotional information, the server generates a dataset for habit removal, which involves applying a generative AI model.

[1881] The server adjusts the speed of fast-paced speech, removes fillers, and adjusts speech speed and volume based on emotional information.

[1882] Input: Data with speaker habits identified and emotional feedback

[1883] Output: Corrected and adjusted audio data

[1884] Step 12: Regenerating spontaneous speech

[1885] The server recreates natural speech based on the corrected and adjusted voice data.

[1886] The server runs the regenerated audio data through a sound quality optimization engine to convert it into a clear, audible format.

[1887] Input: Corrected and adjusted audio data

[1888] Output: Regenerated and optimized audio data

[1889] Step 13: Sending the Regenerated Audio

[1890] The server transmits the regenerated voice data to the terminal in real time.

[1891] Input: Regenerated and optimized audio data

[1892] Output: Audio data sent to the device

[1893] Step 14: Playing back audio data

[1894] The terminal receives the regenerated voice data and plays it back to the user, so that the user hears modified, natural-sounding speech.

[1895] Input: Audio data sent to the device

[1896] Output: The audio played to the user

[1897] (Application example 2)

[1898] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1899] When interacting with customers in real-world situations, customer support staff are required to respond immediately to their customers' emotions and reactions, but they lack the technology to appropriately recognize the speaker's habits and emotions and adjust their speech in real time. This can make natural communication difficult, which can lead to a decrease in customer satisfaction.

[1900] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1901] In this invention, the server includes means for inputting voice data, means for analyzing the input voice data in real time and extracting features of the speaker, means for identifying the speaker's habits based on the extracted features and removing the habits, means for regenerating the voice data from which the habits have been removed as natural speech, means for playing back the regenerated voice data to the user, means for recognizing emotions and adjusting a method for correcting the voice data based on the recognition results, and means for performing voice analysis and emotion recognition when interacting with a customer in real space and supporting a response according to the customer's emotions. This allows voice to be regenerated as natural speech in real time in accordance with the customer's emotions, enabling customer support staff to communicate more smoothly.

[1902] "Sound data" refers to an acoustic signal expressed in audio format.

[1903] "Input means" refers to a hardware or software mechanism for inputting voice data into the system.

[1904] The "analyzing means" is a component that has the function of processing input voice data and extracting specific features.

[1905] "Features" are data extracted from speech data that indicate the speaker's habits and speech characteristics.

[1906] The "means for identifying habits" is a device that has the function of identifying specific phrases, fillers, etc. from a speaker's speech.

[1907] The "means for removing" is a system that has the function of performing processing to remove the habits of the identified speaker.

[1908] The "regenerating means" is a mechanism for generating natural speech based on speech data from which habits have been removed.

[1909] A "means for playing" is a device for providing a user with reproduced natural speech sounds.

[1910] The "means for recognizing emotions" is a component that has the function of identifying emotions from the user's voice data and feeding that information back to the system.

[1911] The "means for adjusting the correction method" is a mechanism having a function for dynamically changing the method for correcting voice data based on the emotion recognition result.

[1912] "Real space" refers to the physical world, the environment in which a user actually exists and acts.

[1913] A "customer" is an entity that uses a service and is the person with whom the service is interacted.

[1914] "Voice analysis during dialogue" is the process of analyzing voice data in real time during a conversation with a customer and understanding its content.

[1915] The "means for emotion recognition" is a mechanism for analyzing and recognizing customer emotions in real time from voice data during a conversation.

[1916] The "means for supporting the response" is a device that has the function of supporting the customer support staff in order to respond appropriately according to the customer's feelings.

[1917] This invention is applied to a customer support system that handles customer support in the real world, and analyzes voice data in real time, removing the speaker's habits and regenerating natural speech. In addition, by using an emotion engine, the content of speech is adjusted based on the customer's emotions.

[1918] System Configuration

[1919] The system consists of the following main components:

[1920] Terminal: A device that inputs voice data and transmits the data to a server. As a concrete example, smart glasses are used.

[1921] Server: Analyzes voice data, extracts features, identifies and removes habits, reproduces natural speech, and recognizes emotions. This uses the Google Speech Recognition API and Transformers library.

[1922] User: The entity that inputs voice data through the system and receives modified, natural-sounding speech. Specifically, customer support staff.

[1923] Hardware configuration and data calculation

[1924] 1. Capture and transmit audio data

[1925] The device captures audio data through a microphone built into the smart glasses, which is temporarily buffered within the device and then transmitted to the server in real time.

[1926] 2. Voice Recognition

[1927] The server converts the voice data received from the device into text using the Google Speech Recognition API, which provides highly accurate voice recognition.

[1928] 3. Emotion recognition

[1929] The server uses an emotion engine (a model using the Transformers library) based on the converted text to identify the customer's emotions, which typically include emotion categories such as joy, anger, and sadness.

[1930] 4. Correcting speech habits

[1931] The server processes the text to identify and remove slurred speech, fillers, rapid speech, and other habits. The removed features are saved as the corrected text.

[1932] 5. Reproducing natural speech

[1933] The server then regenerates the corrected text and produces natural-sounding speech, adjusting the volume and speech rate based on the emotional information obtained from the emotion engine.

[1934] 6. Playing the corrected audio

[1935] The terminal receives the modified voice data sent from the server and plays it back to the user (customer support staff) through the smart glasses' speaker.

[1936] Specific examples

[1937] Physical store scenario

[1938] 1. Customer: "Um, can you tell me what features this product has?"

[1939] 2. The smart glasses capture the customer's speech and send the voice data to a PC.

[1940] 3. The PC receives the voice data and converts it into text using the Google Speech Recognition API.

[1941] 4. The server performs sentiment analysis on the converted text and recognizes the sentiment of "high interest."

[1942] 5. The program corrected the sentence by removing the "um" from "Um, can you tell me what features this product has?"

[1943] 6. Revised text: "Can you tell me what features this product has?"

[1944] 7. The regenerated audio is played back to the customer support staff through the smart glasses' speakers.

[1945] Prompt Sentence Examples

[1946] Question: "Please explain how the program can correct and regenerate speech when a customer says, 'Um, can you tell me what features this product has?'"

[1947] Example input: Customer: "Um, can you tell me what features this product has?"

[1948] Example output: Regenerated speech: "Can you tell me what features this product has?"

[1949] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1950] Step 1:

[1951] Capture and transmit audio data

[1952] A microphone built into the user's smart glasses captures the customer's speech in real time, generating digital audio data. This audio data is temporarily buffered inside the device and then sent to the server via the device's communication function. The input is the customer's speech, and the output is digital audio data sent to the server.

[1953] Step 2:

[1954] Voice Recognition

[1955] The server processes the voice data received from the device and converts it into text using the Google Speech Recognition API. This process results in the content of the voice data being expressed in text form. The input is the voice data sent to the server, and the output is the corresponding text data.

[1956] Step 3:

[1957] emotion recognition

[1958] The server analyzes emotions based on the converted text data using the Transformers library. It uses a generative AI model to analyze emotions from the customer's speech and obtains the results. The input is the converted text data, and the output is the customer's emotional data (e.g., joy, anger, sadness, etc.).

[1959] Step 4:

[1960] Correcting speech habits

[1961] The server identifies and removes speaker habits (fillers, slurred speech, rapid speech, etc.) from the text data along with the emotion recognition results. At this stage, unnecessary fillers and meaningless words are removed from the text. The inputs are the emotion recognition results and the text data, and the output is the corrected text data.

[1962] Step 5:

[1963] Natural speech reproduction

[1964] The server regenerates natural-sounding speech based on the corrected text data and emotion recognition results, adjusting the volume and speech rate. During this process, the text data is converted into clear speech data. The inputs are the corrected text data and emotion recognition results, and the output is the regenerated speech data.

[1965] Step 6:

[1966] Playing the corrected audio

[1967] The terminal receives the regenerated voice data sent from the server and plays it back to the user through the smart glasses speaker. In this process, the customer support staff as the user can hear the corrected natural-sounding speech. The input is the regenerated voice data, and the output is the reproduced voice provided to the user.

[1968] At each step, appropriate data processing or calculation is performed based on specific input data, and the output data required for the next step is generated as a result, thereby realizing natural communication with customers.

[1969] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1970] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1971] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1972] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1973] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1974] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1975] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1976] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1977] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1978] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1979] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1980] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1981] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1982] 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.

[1983] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1984] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1985] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1986] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1987] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1988] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1989] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1990] The following is further disclosed regarding the above embodiment.

[1991] (Claim 1)

[1992] a means for inputting voice data;

[1993] A means for analyzing input voice data in real time and extracting speaker features;

[1994] A means for identifying a speaker's habits based on the extracted features and removing those habits;

[1995] means for regenerating the speech data from which the quirks have been removed as natural speech;

[1996] means for playing the regenerated audio data to the user;

[1997] A system including:

[1998] (Claim 2)

[1999] 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

[2000] (Claim 3)

[2001] The system according to claim 1, characterized in that it identifies speaking habits such as quick speech, stammering, and fillers based on the extracted features.

[2002] "Example 1"

[2003] (Claim 1)

[2004] a means for inputting voice data;

[2005] means for capturing input voice data in real time and transmitting the data to a server;

[2006] means for performing pre-processing of noise reduction and volume adjustment on the received audio data;

[2007] A means for extracting features from the preprocessed speech data and identifying speaker habits;

[2008] applying a generative model to remove the identified habits and regenerate natural-sounding speech; and

[2009] means for transmitting the regenerated audio data to the terminal for playback to the user;

[2010] A system including:

[2011] (Claim 2)

[2012] 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

[2013] (Claim 3)

[2014] The system according to claim 1, characterized in that it identifies speaking habits such as quick speech, stammering, and fillers based on the extracted features.

[2015] "Application Example 1"

[2016] (Claim 1)

[2017] a means for inputting voice data;

[2018] A means for analyzing input voice data in real time and extracting speaker features;

[2019] A means for identifying a speaker's habits based on the extracted features and removing those habits;

[2020] means for regenerating the speech data from which the quirks have been removed as natural speech;

[2021] means for playing the regenerated audio data to a user;

[2022] A system that includes a means for analyzing and correcting the voice of store staff to ensure smooth customer service in a store.

[2023] (Claim 2)

[2024] 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

[2025] (Claim 3)

[2026] The system according to claim 1, characterized in that it identifies speaking habits such as quick speech, stammering, and fillers based on the extracted features.

[2027] "Example 2: Combining Emotion Engines"

[2028] (Claim 1)

[2029] a means for inputting voice data;

[2030] A means for analyzing input voice data in real time;

[2031] A means for extracting features from the analyzed voice data;

[2032] a means for identifying a speaker's habits based on the extracted features and removing the habits;

[2033] means for recognizing a user's emotions in real time and transmitting the information to an analysis means;

[2034] means for regenerating the speech data from which habits have been removed and which has been adjusted based on emotion information as natural speech;

[2035] means for playing the regenerated audio data to the user;

[2036] A system including:

[2037] (Claim 2)

[2038] 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

[2039] (Claim 3)

[2040] The system according to claim 1, characterized in that it identifies speaking habits such as quick speech, stammering, and fillers based on the extracted features.

[2041] (Claim 4)

[2042] 2. The system according to claim 1, wherein the speech rate and volume are adjusted based on the extracted emotional information.

[2043] "Application example 2 when combining emotion engines"

[2044] (Claim 1)

[2045] a means for inputting voice data;

[2046] A means for analyzing input voice data in real time and extracting speaker features;

[2047] A means for identifying a speaker's habits based on the extracted features and removing those habits;

[2048] means for regenerating the speech data from which the quirks have been removed as natural speech;

[2049] means for playing the regenerated audio data to a user;

[2050] means for recognizing emotions and adjusting how the speech data is modified based on the recognition results;

[2051] It is a means of supporting responses that correspond to the customer's emotions by performing voice analysis and emotion recognition when interacting with customers in real space.

[2052] A system including:

[2053] (Claim 2)

[2054] 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

[2055] (Claim 3)

[2056] The system according to claim 1, characterized in that it identifies speaking habits such as quick speech, stammering, and fillers based on the extracted features. [Explanation of symbols]

[2057] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for inputting voice data; A means for analyzing input voice data in real time and extracting speaker features; A means for identifying a speaker's habits based on the extracted features and removing those habits; means for regenerating the speech data from which the quirks have been removed as natural speech; means for playing the regenerated audio data to the user; A system including:

2. 2. The system according to claim 1, wherein noise reduction and volume adjustment are performed as preprocessing of the audio data.

3. 2. The system according to claim 1, wherein speaking habits such as quick speech, stammering, and fillers are identified based on the extracted feature amounts.

Images