system

Abstract

We provide the system. [Solution] A means of receiving audio data of the deceased, A method for extracting speech features from received speech data, A means for training a speech synthesis model using extracted speech features, A means for converting input text into speech data based on a trained speech synthesis model, A means for outputting the generated audio data, A system that includes this.

Description

Technical Field

[0005] ,

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

Background Art

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

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] There is a problem that the voice and words of the deceased fade over time, and the means for family members and friends to share emotions and obtain mental healing through conversations with the deceased are limited. Therefore, there is a need to provide a communication means that can reproduce the voice of the deceased and allow the bereaved to feel the connection with the deceased again.

Means for Solving the Problems

[0005] The present invention provides means for receiving voice data of the deceased, extracting voice features from the data, and training a voice synthesis model. Further, by providing a system that converts input text into voice data using the trained model and outputs the generated voice data, the voice of the deceased is reproduced, providing an experience for the bereaved to converse with the deceased again.

[0006] The term "deceased person" refers to someone who has already passed away, and who remains an important figure to their family and friends.

[0007] "Audio data" refers to digital information that records and stores human speech, and is in a playable data format.

[0008] "Speech features" are specific acoustic properties extracted from speech data, and are elements that indicate the individuality of speech.

[0009] A "speech synthesis model" is an algorithm trained to take text data as input and generate speech data that resembles a human voice.

[0010] "Training" refers to the learning process that a computer model undergoes based on pre-provided data so that it can perform a specific task.

[0011] "Input text" refers to written text information that the user wishes to be played back as audio.

[0012] "Generated audio data" refers to the digital audio output created by a speech synthesis model based on the input text.

[0013] "Outputting" refers to providing processed or generated data in a format usable by the user. [Brief explanation of the drawing]

[0014] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4]It is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] It is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] It is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] It is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] It is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] It shows an emotion map to which a plurality of emotions are mapped. [Figure 10] It shows an emotion map to which a plurality of emotions are mapped. [Figure 11] It is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] It is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] It is a sequence diagram showing the processing flow of the data processing system in Example 2 when an emotion engine is combined. [Figure 14] It is a sequence diagram showing the processing flow of the data processing system in Application Example 2 when an emotion engine is combined.

Embodiments for Carrying Out the Invention

[0015] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0016] <00000​​​​In the following embodiments, the signed processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Furthermore, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), and APU (Accelerated Processing Unit).

[0018] In the following embodiments, signed RAM (Random Access Memory) is a memory that temporarily stores information and is used as work memory by the processor.

[0019] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0020] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0021] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0022] [First Embodiment]

[0023] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0024] As shown in Figure 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0025] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0026] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0027] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0028] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0029] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0030] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0031] As shown in Figure 2, in the data processing device 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0032] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0033] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0034] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0035] This invention is a speech synthesis system that reproduces the voice of a deceased person, enabling a dialogue experience between the bereaved family and the deceased. The system begins with the user uploading the deceased person's voice data to a server. The server processes this voice data, removing noise and cleaning the audio, and then extracts voice features. These obtained voice features are used to train a speech synthesis model, which then builds a model that mimics the deceased person's speaking style and voice tone.

[0036] The user provides input text from their device, which is then sent to the server. Based on the input text received from the user, the server uses a trained speech synthesis model to generate audio data that reproduces the deceased person's voice. This generated audio data is then sent back to the user's device and delivered to the user through the device in the deceased person's voice.

[0037] As a concrete example, if a user enters the text "Please tell me how your day was today," the server processes the text using a speech synthesis model and generates audio data that mimics the deceased person's voice. This audio data is sent to the terminal, and the user receives the message in the deceased person's voice, experiencing the feeling as if the deceased person is speaking to them in person.

[0038] In this way, the present invention aims to provide a new connection with the deceased and bring healing to the hearts of bereaved families. This system allows users to look back on past memories and once again have the experience of interacting with the deceased through their voice.

[0039] The following describes the processing flow.

[0040] Step 1:

[0041] Users collect audio data of deceased individuals and upload it to a server via their devices. This includes audio files extracted from past recordings and video clips.

[0042] Step 2:

[0043] The server analyzes the received audio data. First, it performs audio preprocessing such as noise reduction and echo reduction to improve sound quality.

[0044] Step 3:

[0045] The server extracts speech features from the pre-processed audio data. This involves a process of calculating features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants.

[0046] Step 4:

[0047] The server uses the extracted speech features to train a speech synthesis model. This allows the model to acquire the ability to mimic the deceased person's speaking style and tone of voice.

[0048] Step 5:

[0049] The user sends the text they want to reproduce in the deceased person's voice as input text to the server via their device.

[0050] Step 6:

[0051] The server passes the received input text to a speech synthesis model, which generates audio data that mimics the deceased person's voice. In this process, the model synthesizes the input text as an audio waveform.

[0052] Step 7:

[0053] The generated audio data is sent from the server to the user's device. The device then converts this audio data into a playable format.

[0054] Step 8:

[0055] The device plays audio data generated through speakers or headphones to the user. The user listens to messages conveyed in the deceased's voice, experiencing a feeling as if they are conversing with the deceased again.

[0056] (Example 1)

[0057] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0058] Currently, many bereaved families wish to feel a connection with their loved ones, but there are limited ways to communicate using the deceased's voice. Existing voice reproduction technologies struggle to realistically reproduce the voice of the deceased, and are insufficient to help bereaved families reminisce about the past. Therefore, there is a growing expectation for technologies that faithfully reproduce the voice of the deceased, providing an experience as if one were conversing with them.

[0059] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0060] In this invention, the server includes means for receiving audio information, means for analyzing the characteristics of the audio from the received audio information, and means for training an information conversion model using the analyzed characteristics of the audio. This allows for the faithful reproduction of the deceased's voice, enabling the user to relive a conversation with the deceased.

[0061] "Audio information" refers to digital or analog information formats used to represent sound, such as audio data or audio signals.

[0062] "Means of receiving" refers to the technology or device used to receive data transmitted from an external source.

[0063] "Means of analysis" refers to techniques or devices used to analyze information and extract specific features or patterns.

[0064] "Features" refer to the characteristics and attributes of the information being analyzed, and are indicators of the data extracted during the analysis process.

[0065] An "information conversion model" is an algorithm or program used to convert input data into another data format.

[0066] "Training methods" refer to the process of improving the performance of a model or algorithm using data to achieve desired results.

[0067] "Inputted character information" refers to information entered as character data, primarily data expressed in text format.

[0068] "Means of conversion" refers to the technology or process of changing data in one format to another.

[0069] "Means of output" refers to technology or equipment for providing processed data to an external party.

[0070] "Means for removing noise" refers to techniques or devices for removing unwanted acoustic components from audio information.

[0071] "User's device" refers to electronic devices and terminals used by the user, including those capable of receiving, displaying, or playing back data.

[0072] This invention relates to a speech synthesis system that reproduces the voice of a deceased person and provides bereaved families with an experience of interacting with the deceased. This system is primarily built through the interaction of a server, a terminal, and the user.

[0073] The user uploads the deceased person's voice data to the server via their device. The server uses an audio editing library (e.g., librosa) to remove noise from the data and improve the sound quality. Then, it uses an audio analysis tool (e.g., Praat) to extract the voice's characteristics. This results in the acquisition of data such as the voice's pitch, range, and timbre.

[0074] Next, the server trains a speech synthesis model. This model uses a machine learning framework (e.g., Tensorflow®, PyTorch). The training is performed to mimic the deceased person's unique speaking style and tone of voice based on the extracted speech features. This creates an information transformation model that can naturally reproduce the deceased person's voice.

[0075] The user enters a message they wish to convey to the deceased as text into the terminal. When this text is sent to the server, the server uses a trained model to convert the text into speech. During this process, the prompt "Please mimic everyday conversation in the voice of the deceased" is taken into consideration.

[0076] As a concrete example, if a user enters the text "Please tell me how your day was," the server converts this into audio data. The generated audio data is reproduced in the deceased person's voice and provided to the user through their device. This process allows the user to feel as if they are conversing with the deceased again and to immerse themselves in past memories through that voice.

[0077] This system aims to help bereaved families rebuild their connection with the deceased and find emotional healing.

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

[0079] Step 1:

[0080] The user uploads the deceased person's voice data to the server using their device. The input consists of the deceased person's voice files (e.g., WAV, MP3 format). The server receives these files and prepares for subsequent audio processing.

[0081] Step 2:

[0082] The server removes noise and cleans the received audio data. The input is the uploaded audio data, and background noise and unwanted sounds are removed using a noise reduction library (e.g., librosa). The output is cleaned, high-quality audio data. Specific operations include filtering.

[0083] Step 3:

[0084] The server extracts audio features from the cleaned audio data. The input is processed audio data, and an audio analysis tool (e.g., Praat) is used to extract numerical characteristics such as pitch, range, and intensity. The output is an audio feature vector. This process involves specific frequency analysis and time-series analysis.

[0085] Step 4:

[0086] The server trains a speech synthesis model. The input is a vector of speech features, and a speech synthesis model capable of reproducing the voice of a deceased person is built through a machine learning framework (e.g., TensorFlow, PyTorch). The output is the trained speech synthesis model. Specifically, model parameters are optimized and adjusted through a feedback loop.

[0087] Step 5:

[0088] The user sends input text to the server via their device. The server receives text information specified by the user as input (e.g., "Please tell me how your day was today").

[0089] Step 6:

[0090] The server uses a trained speech synthesis model to convert input text into speech data. The input consists of text from the user and the trained model, and the generative AI model converts the text data into speech data. Here, the prompt "Imitate an everyday conversation in the voice of the deceased" is considered. The output is speech data that imitates the voice of the deceased.

[0091] Step 7:

[0092] The server sends the generated audio data to the user's device. The input is the generated audio data, which the user's device receives. The device then uses an audio player or similar device to play the deceased person's voice for the user. The output is the audio message that reaches the user.

[0093] (Application Example 1)

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

[0095] Maintaining memories and the experience of interacting with a deceased loved one through their voice is a means of spiritual healing for many, but in reality, the means to achieve this are very limited. In particular, dynamically reproducing the voice of a deceased person in response to the surrounding environment or specific situations has been difficult with conventional technology. In response to this, there is a need for technology that allows users to experience the voice of a deceased person more realistically in a virtual environment, thereby allowing them to feel past memories more deeply.

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

[0097] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, means for converting input text into voice data based on a trained speech synthesis model, means for detecting location information for playing the deceased's voice within a virtual environment, and means for dynamically generating and playing the deceased's voice data according to specific circumstances within the virtual environment. This enables the user to have an interactive experience using the voice of the deceased within the virtual environment.

[0098] "Voice data of a deceased person" refers to a collection of audio signals in which the voice of a deceased person has been recorded and saved.

[0099] "Speech features" are data extracted from acoustic signals that represent the nature and phonetic characteristics of a voice.

[0100] A "speech synthesis model" is an algorithm or structure for generating speech data that mimics a human voice based on speech features.

[0101] "Input text" refers to a string of characters that a user manually or automatically enters to give instructions or information to a computer.

[0102] "Audio data" refers to a collection of information recorded from sound and stored in digital or analog format.

[0103] A "virtual environment" is a space or situation artificially constructed using computer technology, providing users with experiences that do not exist in reality.

[0104] "Location information" refers to data that indicates the physical location of a specific object or user, and is usually expressed in coordinate format.

[0105] "Means of dynamically generating and reproducing information" refers to methods and processes for generating and providing information to users in real time, while changing it according to the situation and conditions.

[0106] The system that realizes this invention includes a series of processes to reproduce the voice of a deceased person and allow the user to experience it in a virtual environment. The server receives the deceased person's voice data and processes it to extract voice features. This includes data preprocessing using a noise reduction algorithm to clean up the voice data. Acoustic signal analysis software is used to extract voice features. Subsequently, machine learning libraries such as PyTorch are used to train a speech synthesis model. This model can reproduce the deceased person's voice based on the extracted voice features.

[0107] The server then generates audio data using a speech synthesis model based on the text entered by the user. This generation process is performed using the Google® Cloud Text-to-Speech API, and the speech conversion is performed in real time. The generated audio data is sent to the user's smart glasses or head-mounted display. This allows the user to hear the voice of a deceased person based on specific location information while moving around in the virtual environment.

[0108] By utilizing sensors built into the device, the system acquires the user's location information in real time and dynamically plays audio according to the location. For this purpose, Firebase is used for information management, and audio messages associated with the user's location and selected virtual items are provided.

[0109] As a concrete example, when a user reaches a specific store in a virtual shopping mall, a voice recording of the deceased reminiscing about their favorite items is played. An example of a prompt message would be: "Create a message using the deceased's voice, telling the user about Christmas memories, and adding specific anecdotes that evoke warm family moments." This allows the user to have an experience as if they were having a conversation with the deceased.

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

[0111] Step 1:

[0112] The server receives audio data of the deceased from the user. This input data is preprocessed to verify the consistency of its format and sample rate, and then saved as an audio file.

[0113] Step 2:

[0114] The server extracts speech features from the received audio data. Here, speech signal processing techniques are used to extract Mel-frequency cepstrum coefficients (MFCCs), among other things. This data processing is intended to numerically represent the characteristics of the speech data, and the results are output as a list of speech features.

[0115] Step 3:

[0116] The server trains a speech synthesis model using speech features. Using machine learning libraries (e.g., PyTorch), it constructs models such as multilayer perceptrons and recurrent neural networks using the speech features as input, and then adjusts the model parameters. The output is a trained model capable of mimicking the voice of a deceased person.

[0117] Step 4:

[0118] The user enters the input text to be used in the virtual environment into the terminal. This text is sent to the server as the message the user wants the deceased to say.

[0119] Step 5:

[0120] The server uses a trained speech synthesis model based on the received text to generate speech data. The Google Cloud Text-to-Speech API is used to convert the input text into speech data and generate it as data.

[0121] Step 6:

[0122] The server sends the generated audio data to the user's device. The device receives this audio data and prepares to play it back in the deceased person's voice.

[0123] Step 7:

[0124] The user's device uses its built-in sensors to detect its position within the virtual environment in real time. This location information is output as location coordinates because it affects the timing of audio data playback.

[0125] Step 8:

[0126] Based on its location, the device dynamically plays audio data received from the server when it reaches a specific point within the virtual environment. Specifically, a message in the deceased's voice, linked to the location, is played, allowing the user to experience a conversation with the deceased.

[0127] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0128] This invention combines a system for reproducing the voice of a deceased person with an emotion engine that recognizes the user's emotions. The system begins by uploading the deceased person's voice data to a server. The server removes noise from the received voice data and extracts voice features. This results in higher quality voice data, which is used to train a speech synthesis model.

[0129] The user inputs text they wish to be reproduced in the deceased person's voice through their device, while an emotion engine analyzes the user's emotional state. The emotion engine analyzes the input text, voice intonation, volume, tempo, etc., to determine the user's emotions. This analysis result is fed back to the speech synthesis model, influencing the generated voice data. As a result, the generated voice is in harmony with the user's emotions, creating a more personalized conversational experience.

[0130] For example, if a user types "Today was a good day," the emotion engine analyzes the user's emotion as joy and sends that information to the speech synthesis model. Based on this information, the server generates a voice with a softer, more cheerful tone and sends it to the user's device. The device plays the voice, and the user can experience what it's like to share joy with a loved one through their voice.

[0131] Based on this format, users can have an emotional connection with the deceased through recreated voices, allowing them to find solace. Furthermore, the introduction of an emotion engine makes the conversational experience more personalized, providing communication that is attentive to the user's emotions at any given moment.

[0132] The following describes the processing flow.

[0133] Step 1:

[0134] Users collect audio data of deceased individuals and upload it to the server. This includes audio clips and audio files extracted from videos.

[0135] Step 2:

[0136] The server receives the uploaded audio data and removes noise from it. It analyzes the audio to reduce unwanted background noise and performs pre-processing to improve sound quality.

[0137] Step 3:

[0138] From the processed audio data, the server extracts audio features. Specifically, it calculates acoustic features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants to obtain the deceased person's characteristic voice pattern.

[0139] Step 4:

[0140] The server trains a speech synthesis model based on the extracted speech features. This training enables the model to accurately mimic the deceased person's voice quality and speaking style.

[0141] Step 5:

[0142] Users input text they want to reproduce in the deceased's voice via their device and send it to the server. The text they input is what they wanted to exchange with the deceased or any messages they wished they had.

[0143] Step 6:

[0144] The emotion engine operates on the server, analyzing the user's input text and their current voice or input method to recognize their emotional state. This analysis determines whether the user is expressing a specific emotion, such as joy, sadness, or excitement.

[0145] Step 7:

[0146] Based on the analysis results, the server passes the input text to a speech synthesis model and generates audio data by applying emotional feedback. This adjusts the deceased person's voice to match the user's emotions in terms of tone and intonation.

[0147] Step 8:

[0148] The generated audio data is sent from the server to the user's terminal. The terminal converts the received audio data into a playable format and provides it to the user.

[0149] Step 9:

[0150] The device plays audio generated through speakers or earphones to the user. By listening to this personalized audio, the user can feel an emotional connection, as if the deceased person were present.

[0151] (Example 2)

[0152] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0153] In modern society, there is a need for ways to more intimately recreate memories of the deceased and maintain an emotional connection. In particular, providing users with the opportunity to find emotional healing through the voice of the deceased is crucial. However, existing technologies have limitations in adapting the voice of the deceased to their emotions and improving voice quality. Furthermore, the lack of voice generation methods that consider the user's emotional state makes it difficult to provide a personalized conversational experience.

[0154] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0155] In this invention, the server includes means for receiving voice information of the deceased, means for extracting acoustic features from the received voice information, means for reducing noise from the acoustic features, means for evaluating the user's emotional state using an emotion analysis engine, means for reflecting the evaluated emotional state in a generation model and converting the information into speech, and means for distributing the generated speech to the user's terminal. This makes it possible to reproduce the voice of the deceased in high quality and in a manner adapted to the user's emotions, thereby providing an individualized emotional dialogue experience.

[0156] "Voice information of the deceased" refers to audio data of an individual recorded in the past, and this data serves as the basic information for analyzing and reconstructing the voice of the deceased.

[0157] "Acoustic features" are numerical indicators that show the attributes of speech, such as Mel-frequency cepstrum coefficients, pitch, and waveform data extracted from speech data.

[0158] "Methods for reducing noise" refer to processes that remove unwanted noise from audio data and improve sound quality.

[0159] A "generative model" is a model trained using machine learning techniques, and it is an algorithm that generates speech based on text and emotional states.

[0160] An "emotion analysis engine" is software or a system that evaluates a user's emotional state based on the characteristics of input text or audio.

[0161] "User's emotional state" refers to the emotional state a user exhibits when using voice input or text input, and includes emotions such as joy, sadness, and anger.

[0162] "User's device" refers to a device used to receive and play back generated audio data, and includes smartphones, personal computers, and other similar devices.

[0163] This invention provides a more personalized conversational experience in a system that reproduces the voice of a deceased person by taking into account the user's emotional state. Details for implementing this system are provided below.

[0164] 1. Processing of audio data:

[0165] The user uploads the deceased person's voice data to the server via their device. The server uses voice processing software (e.g., Python libraries librosa and pydub) to extract acoustic features. This step involves a noise reduction process to improve the quality of the voice data.

[0166] 2. Emotion analysis and speech synthesis:

[0167] The server uses a generative AI model (e.g., Tacotron 2 or WaveNet) to train a speech synthesis model based on extracted acoustic features. When the user inputs text they want to reproduce in the deceased person's voice from their device, their emotional state is analyzed through an emotion analysis engine (e.g., a natural language processing API). This analysis result is provided to the generative AI model as a prompt along with the text information.

[0168] 3. Speech generation and output:

[0169] The generative AI model generates speech that reflects the user's emotional state based on the user's input text. The generated speech data is sent from the server to the user's device and played back on the device. Through this process, the user can experience an emotional connection with the voice of a deceased loved one.

[0170] Specific example:

[0171] For example, if a user types "Today was a good day," the sentiment analysis engine interprets this as joy and feeds it back to the generative model. Based on this information, the server generates speech in a soft, cheerful tone. An example of a prompt is: "The user's emotion is joy. Please reproduce 'Today was a good day' in the voice of the deceased."

[0172] This embodiment allows users to enjoy an emotionally resonant audio experience. The system aims to provide users with emotional comfort by offering an advanced audio playback method using the voice of a deceased person.

[0173] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0174] Step 1:

[0175] The user uploads the deceased person's voice data to the server via their device. The input voice data file is selected on the device and sent to the server. The user's specific operation involves selecting the voice file on the file selection screen and pressing the "Upload" button. The output is the voice data saved on the server.

[0176] Step 2:

[0177] The server extracts acoustic features from the received audio data. This process uses audio analysis software (e.g., the librosa library in Python) to perform noise reduction and calculate Mel-frequency cepstrum coefficients (MFCCs). The input is the audio data obtained in step 1, and the output is acoustic feature data with reduced noise.

[0178] Step 3:

[0179] The server sends text data entered from the terminal to the sentiment analysis engine in order to analyze the user's emotions. The text entered by the user on the terminal is subject to sentiment analysis. Accordingly, a natural language processing API is used to determine the emotional state of the input text. The output is data indicating the analyzed emotional state.

[0180] Step 4:

[0181] The server uses emotion data and acoustic features obtained from the emotion analysis engine to perform speech synthesis using a generative AI model. At this stage, the output data from steps 2 and 3 is used as input. The speech synthesis model receives this data as prompts and generates speech appropriate to the emotion. The output is the generated speech data.

[0182] Step 5:

[0183] The server sends the generated audio data to the user's terminal. In this process, the audio data obtained in step 4 is used as input and transferred to the terminal. The output is the audio data received by the terminal.

[0184] Step 6:

[0185] The device plays the received audio data to the user. This process utilizes the device's audio player, and the generated audio is played back to the user through speakers or headphones. This allows the user to have an interactive experience through the voice of the deceased.

[0186] (Application Example 2)

[0187] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0188] In modern times, it is a crucial issue for people to preserve memories of deceased loved ones through audio recordings, thereby gaining emotional connection and healing. While conventional technology could reproduce the voice of a deceased person, it struggled to generate personalized audio that responded to the user's emotions. Therefore, there is a need to provide a system that reflects the user's feelings and reproduces the voice of a deceased person in a more natural and emotionally resonant way.

[0189] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0190] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, and an emotion analysis function for analyzing the user's emotional state. This makes it possible to generate and provide to the user the voice of the deceased after making adjustments according to the user's emotions.

[0191] "Audio data" refers to information that digitizes the waveform of sound, and includes the specific characteristics of the deceased person's voice.

[0192] "Speech features" are individual attributes such as pitch, intensity, and texture extracted from speech data, and these features make it possible to analyze speech patterns.

[0193] A "speech synthesis algorithm" is a computational method for generating digital data that sounds like a human voice based on speech characteristics.

[0194] "Emotion analysis function" is a technology that automatically identifies emotions from user input or voice and quantifies the user's emotional state.

[0195] A "user device" is an electronic device owned by the user, which is a terminal for receiving and playing back generated audio data.

[0196] To implement this invention, a specific speech synthesis system is used. First, the server receives the deceased person's voice data from the user. This voice data is processed on the server, first undergoing noise reduction. Next, voice features are extracted. These voice features include the basic pitch, timbre, and prosody of the deceased person's voice.

[0197] The server trains a speech synthesis algorithm using the extracted speech features. This algorithm could, for example, be a Transformer-based speech synthesis model. The trained model can then accept text provided by the user as input and generate speech data in the deceased person's voice.

[0198] The user terminal has the function of receiving and playing back the generated audio data. Furthermore, an emotion analysis function is introduced to analyze the user's emotions. Emotion analysis can be performed from the user's input text, voice intonation, and facial expression data acquired from the camera. This uses software such as the Python library "librosa" and "OpenCV".

[0199] The results of the emotion analysis are fed back to the speech synthesis model and reflected in the generated speech. For example, if a user says "Today was a good day" to their smartphone, that emotion is analyzed as joy, and the server uses this information to generate speech in a cheerful tone.

[0200] An example of a prompt to input into a generative AI model is the instruction, "The user is requesting X, and their emotion is estimated to be Y. Provide advice in the voice of the deceased that is in harmony with the emotion." This allows for a more personalized voice experience that is attentive to the user's emotional state.

[0201] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0202] Step 1:

[0203] The user uses a smartphone or other device to input the text they want to be reproduced in the deceased person's voice. The entered text is sent from the user's device to the server. At this point, the input is text data and is saved as data ready to be sent to the server as output.

[0204] Step 2:

[0205] The server receives audio data of the deceased person, provided in advance by the user. Since this audio data may contain noise, the server applies a noise reduction algorithm to remove unwanted sounds. The input is the audio data of the deceased person, and the output is the audio data with the noise removed.

[0206] Step 3:

[0207] The server extracts speech features from the denoised audio data. These speech features include pitch, tempo, and voice quality. This process uses signal processing techniques to quantify these features. The input is denoised audio data, and the output is the extracted speech features.

[0208] Step 4:

[0209] The server trains a speech synthesis algorithm using the extracted speech features. Deep learning techniques are used to enhance the algorithm so that it can reproduce the deceased person's speech patterns. The input is speech features, and the output is the trained speech synthesis model.

[0210] Step 5:

[0211] In parallel, the user terminal uses input text, audio, and video data obtained from the camera to analyze the user's emotions. Using tools such as Python's "librosa" and "OpenCV," the user's intonation and facial expressions are analyzed, and their emotional state is quantified. The input is text, audio, and video data, and the output is the user's emotional parameters.

[0212] Step 6:

[0213] The server feeds the user's emotion parameters back into the speech synthesis model, which then converts the input text into speech in the deceased person's voice. At this stage, the speech is generated with a tone and tempo that reflects the user's emotions. The input consists of the user's emotion parameters and text, and the output is emotion-harmonized speech data.

[0214] Step 7:

[0215] The generated audio data is sent from the server to the user's terminal, which then plays the audio. Through this output, the user can experience what it's like to converse with the deceased person. The input is the generated audio data, and the output is the audio itself that is played back.

[0216] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0217] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0218] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0219] [Second Embodiment]

[0220] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0221] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0222] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0223] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0224] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0225] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0226] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0227] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0228] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0229] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0230] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0231] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0232] This invention is a speech synthesis system that reproduces the voice of a deceased person, enabling a dialogue experience between the bereaved family and the deceased. The system begins with the user uploading the deceased person's voice data to a server. The server processes this voice data, removing noise and cleaning the audio, and then extracts voice features. These obtained voice features are used to train a speech synthesis model, which then builds a model that mimics the deceased person's speaking style and voice tone.

[0233] The user provides input text from their device, which is then sent to the server. Based on the input text received from the user, the server uses a trained speech synthesis model to generate audio data that reproduces the deceased person's voice. This generated audio data is then sent back to the user's device and delivered to the user through the device in the deceased person's voice.

[0234] As a concrete example, if a user enters the text "Please tell me how your day was today," the server processes the text using a speech synthesis model and generates audio data that mimics the deceased person's voice. This audio data is sent to the terminal, and the user receives the message in the deceased person's voice, experiencing the feeling as if the deceased person is speaking to them in person.

[0235] In this way, the present invention aims to provide a new connection with the deceased and bring healing to the hearts of bereaved families. This system allows users to look back on past memories and once again have the experience of interacting with the deceased through their voice.

[0236] The following describes the processing flow.

[0237] Step 1:

[0238] Users collect audio data of deceased individuals and upload it to a server via their devices. This includes audio files extracted from past recordings and video clips.

[0239] Step 2:

[0240] The server analyzes the received audio data. First, it performs audio preprocessing such as noise reduction and echo reduction to improve sound quality.

[0241] Step 3:

[0242] The server extracts speech features from the pre-processed audio data. This involves a process of calculating features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants.

[0243] Step 4:

[0244] The server uses the extracted speech features to train a speech synthesis model. This allows the model to acquire the ability to mimic the deceased person's speaking style and tone of voice.

[0245] Step 5:

[0246] The user sends the text they want to reproduce in the deceased person's voice as input text to the server via their device.

[0247] Step 6:

[0248] The server passes the received input text to a speech synthesis model, which generates audio data that mimics the deceased person's voice. In this process, the model synthesizes the input text as an audio waveform.

[0249] Step 7:

[0250] The generated audio data is sent from the server to the user's device. The device then converts this audio data into a playable format.

[0251] Step 8:

[0252] The device plays audio data generated through speakers or headphones to the user. The user listens to messages conveyed in the deceased's voice, experiencing a feeling as if they are conversing with the deceased again.

[0253] (Example 1)

[0254] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0255] Currently, many bereaved families wish to feel a connection with their loved ones, but there are limited ways to communicate using the deceased's voice. Existing voice reproduction technologies struggle to realistically reproduce the voice of the deceased, and are insufficient to help bereaved families reminisce about the past. Therefore, there is a growing expectation for technologies that faithfully reproduce the voice of the deceased, providing an experience as if one were conversing with them.

[0256] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0257] In this invention, the server includes means for receiving audio information, means for analyzing the characteristics of the audio from the received audio information, and means for training an information conversion model using the analyzed characteristics of the audio. This allows for the faithful reproduction of the deceased's voice, enabling the user to relive a conversation with the deceased.

[0258] "Audio information" refers to digital or analog information formats used to represent sound, such as audio data or audio signals.

[0259] "Means of receiving" refers to the technology or device used to receive data transmitted from an external source.

[0260] "Means of analysis" refers to techniques or devices used to analyze information and extract specific features or patterns.

[0261] "Features" refer to the characteristics and attributes of the information being analyzed, and are indicators of the data extracted during the analysis process.

[0262] An "information conversion model" is an algorithm or program used to convert input data into another data format.

[0263] "Training methods" refer to the process of improving the performance of a model or algorithm using data to achieve desired results.

[0264] "Inputted character information" refers to information entered as character data, primarily data expressed in text format.

[0265] "Means of conversion" refers to the technology or process of changing data in one format to another.

[0266] "Means of output" refers to technology or equipment for providing processed data to an external party.

[0267] "Means for removing noise" refers to techniques or devices for removing unwanted acoustic components from audio information.

[0268] "User's device" refers to electronic devices and terminals used by the user, including those capable of receiving, displaying, or playing back data.

[0269] This invention relates to a speech synthesis system that reproduces the voice of a deceased person and provides bereaved families with an experience of interacting with the deceased. This system is primarily built through the interaction of a server, a terminal, and the user.

[0270] The user uploads the deceased person's voice data to the server via their device. The server uses an audio editing library (e.g., librosa) to remove noise from the data and improve the sound quality. Then, it uses an audio analysis tool (e.g., Praat) to extract the voice's characteristics. This results in the acquisition of data such as the voice's pitch, range, and timbre.

[0271] Next, the server trains a speech synthesis model. This model uses a machine learning framework (e.g., TensorFlow, PyTorch). The training is performed to mimic the deceased person's unique speaking style and tone of voice based on the extracted speech features. This creates an information transformation model that can naturally reproduce the deceased person's voice.

[0272] The user enters a message they wish to convey to the deceased as text into the terminal. When this text is sent to the server, the server uses a trained model to convert the text into speech. During this process, the prompt "Please mimic everyday conversation in the voice of the deceased" is taken into consideration.

[0273] As a concrete example, if a user enters the text "Please tell me how your day was," the server converts this into audio data. The generated audio data is reproduced in the deceased person's voice and provided to the user through their device. This process allows the user to feel as if they are conversing with the deceased again and to immerse themselves in past memories through that voice.

[0274] This system aims to help bereaved families rebuild their connection with the deceased and find emotional healing.

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

[0276] Step 1:

[0277] The user uploads the deceased person's voice data to the server using their device. The input consists of the deceased person's voice files (e.g., WAV, MP3 format). The server receives these files and prepares for subsequent audio processing.

[0278] Step 2:

[0279] The server removes noise and cleans the received audio data. The input is the uploaded audio data, and background noise and unwanted sounds are removed using a noise reduction library (e.g., librosa). The output is cleaned, high-quality audio data. Specific operations include filtering.

[0280] Step 3:

[0281] The server extracts the voice features of the cleaned voice data. The input is the processed voice data, and using a voice analysis tool (e.g., Praat), features such as the pitch, range, and intensity of the voice are extracted as numerical values. The output is the voice feature vector. In this process, specific frequency analysis and time series analysis are performed.

[0282] Step 4:

[0283] The server trains a voice synthesis model. The input is the vector of voice features, and through a machine learning framework (e.g., TensorFlow, PyTorch), a voice synthesis model capable of reproducing the voice of the deceased is constructed. The output is the trained voice synthesis model. Specifically, optimization of model parameters and adjustment through a feedback loop are performed.

[0284] Step 5:

[0285] The user sends input text to the server through the terminal. As input, the character information specified by the user (e.g., "Please tell me what kind of day it was today") is sent to the server.

[0286] Step 6:

[0287] The server uses the trained voice synthesis model to convert the input text into voice data. The input is the text from the user and the trained model, and the generative AI model converts the character data into voice data. Here, the prompt sentence "Please imitate a daily conversation in the voice of the deceased" is considered. The output is the voice data imitating the voice color of the deceased.

[0288] Step 7:

[0289] The server sends the generated voice data to the user's terminal. The input is the generated voice data, and the user's terminal receives this data. The terminal uses a voice player or the like to play the voice of the deceased for the user. The output is the voice message reaching the user.

[0290] (Application Example 1)

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

[0292] Maintaining memories and the experience of interacting with a deceased loved one through their voice is a means of spiritual healing for many, but in reality, the means to achieve this are very limited. In particular, dynamically reproducing the voice of a deceased person in response to the surrounding environment or specific situations has been difficult with conventional technology. In response to this, there is a need for technology that allows users to experience the voice of a deceased person more realistically in a virtual environment, thereby allowing them to feel past memories more deeply.

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

[0294] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, means for converting input text into voice data based on a trained speech synthesis model, means for detecting location information for playing the deceased's voice within a virtual environment, and means for dynamically generating and playing the deceased's voice data according to specific circumstances within the virtual environment. This enables the user to have an interactive experience using the voice of the deceased within the virtual environment.

[0295] "Voice data of a deceased person" refers to a collection of audio signals in which the voice of a deceased person has been recorded and saved.

[0296] "Speech features" are data extracted from acoustic signals that represent the nature and phonetic characteristics of a voice.

[0297] A "speech synthesis model" is an algorithm or structure for generating speech data that mimics a human voice based on speech features.

[0298] "Input text" refers to a string of characters that a user manually or automatically enters to give instructions or information to a computer.

[0299] "Audio data" refers to a collection of information recorded from sound and stored in digital or analog format.

[0300] A "virtual environment" is a space or situation artificially constructed using computer technology, providing users with experiences that do not exist in reality.

[0301] "Location information" refers to data that indicates the physical location of a specific object or user, and is usually expressed in coordinate format.

[0302] "Means of dynamically generating and reproducing information" refers to methods and processes for generating and providing information to users in real time, while changing it according to the situation and conditions.

[0303] The system that realizes this invention includes a series of processes to reproduce the voice of a deceased person and allow the user to experience it in a virtual environment. The server receives the deceased person's voice data and processes it to extract voice features. This includes data preprocessing using a noise reduction algorithm to clean up the voice data. Acoustic signal analysis software is used to extract voice features. Subsequently, machine learning libraries such as PyTorch are used to train a speech synthesis model. This model can reproduce the deceased person's voice based on the extracted voice features.

[0304] The server then generates audio data using a speech synthesis model based on the text entered by the user. This generation process is performed using the Google Cloud Text-to-Speech API, and the speech conversion is carried out in real time. The generated audio data is sent to the user's smart glasses or head-mounted display. This allows the user to hear the voice of the deceased based on specific location information while moving around in the virtual environment.

[0305] Utilize the sensors built into the terminal to obtain the user's location information in real time and dynamically play back voices according to the location. For this purpose, Firebase is used for information management, and voice messages associated with the user's location and selected virtual items are provided.

[0306] As a specific example, when the user arrives at a specific store in a virtual shopping mall, a recollection story about the favorite products of a deceased person is played back in that voice. Examples of "prompt sentences" include "Create a message that uses the voice of the deceased person to talk to the user about Christmas memories. Add specific episodes that remind the user of warm times with the family." By doing so, the user can experience as if they are talking to the deceased person.

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

[0308] Step 1:

[0309] The server receives the voice data of the deceased person from the user. This input data is preprocessed to check the consistency of the format and sample rate and is saved as an audio file.

[0310] Step 2:

[0311] The server extracts voice features from the received voice data. Here, techniques of voice signal processing are used to extract Mel-frequency cepstral coefficients (MFCC), etc. This data processing is for numerically expressing the characteristics of the voice data and is output as a list of voice features.

[0312] Step 3:

[0313] The server trains a speech synthesis model using speech features. Using machine learning libraries (e.g., PyTorch), it constructs models such as multilayer perceptrons and recurrent neural networks using the speech features as input, and then adjusts the model parameters. The output is a trained model capable of mimicking the voice of a deceased person.

[0314] Step 4:

[0315] The user enters the input text to be used in the virtual environment into the terminal. This text is sent to the server as the message the user wants the deceased to say.

[0316] Step 5:

[0317] The server uses a trained speech synthesis model based on the received text to generate speech data. The Google Cloud Text-to-Speech API is used to convert the input text into speech data and generate it as data.

[0318] Step 6:

[0319] The server sends the generated audio data to the user's device. The device receives this audio data and prepares to play it back in the deceased person's voice.

[0320] Step 7:

[0321] The user's device uses its built-in sensors to detect its position within the virtual environment in real time. This location information is output as location coordinates because it affects the timing of audio data playback.

[0322] Step 8:

[0323] Based on its location, the device dynamically plays audio data received from the server when it reaches a specific point within the virtual environment. Specifically, a message in the deceased's voice, linked to the location, is played, allowing the user to experience a conversation with the deceased.

[0324] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0325] This invention combines a system for reproducing the voice of a deceased person with an emotion engine that recognizes the user's emotions. The system begins by uploading the deceased person's voice data to a server. The server removes noise from the received voice data and extracts voice features. This results in higher quality voice data, which is used to train a speech synthesis model.

[0326] The user inputs text they wish to be reproduced in the deceased person's voice through their device, while an emotion engine analyzes the user's emotional state. The emotion engine analyzes the input text, voice intonation, volume, tempo, etc., to determine the user's emotions. This analysis result is fed back to the speech synthesis model, influencing the generated voice data. As a result, the generated voice is in harmony with the user's emotions, creating a more personalized conversational experience.

[0327] For example, if a user types "Today was a good day," the emotion engine analyzes the user's emotion as joy and sends that information to the speech synthesis model. Based on this information, the server generates a voice with a softer, more cheerful tone and sends it to the user's device. The device plays the voice, and the user can experience what it's like to share joy with a loved one through their voice.

[0328] Based on this format, users can have an emotional connection with the deceased through recreated voices, allowing them to find solace. Furthermore, the introduction of an emotion engine makes the conversational experience more personalized, providing communication that is attentive to the user's emotions at any given moment.

[0329] The following describes the processing flow.

[0330] Step 1:

[0331] Users collect audio data of deceased individuals and upload it to the server. This includes audio clips and audio files extracted from videos.

[0332] Step 2:

[0333] The server receives the uploaded audio data and removes noise from it. It analyzes the audio to reduce unwanted background noise and performs pre-processing to improve sound quality.

[0334] Step 3:

[0335] From the processed audio data, the server extracts audio features. Specifically, it calculates acoustic features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants to obtain the deceased person's characteristic voice pattern.

[0336] Step 4:

[0337] The server trains a speech synthesis model based on the extracted speech features. This training enables the model to accurately mimic the deceased person's voice quality and speaking style.

[0338] Step 5:

[0339] Users input text they want to reproduce in the deceased's voice via their device and send it to the server. The text they input is what they wanted to exchange with the deceased or any messages they wished they had.

[0340] Step 6:

[0341] The emotion engine operates on the server, analyzing the user's input text and their current voice or input method to recognize their emotional state. This analysis determines whether the user is expressing a specific emotion, such as joy, sadness, or excitement.

[0342] Step 7:

[0343] Based on the analysis results, the server passes the input text to a speech synthesis model and generates audio data by applying emotional feedback. This adjusts the deceased person's voice to match the user's emotions in terms of tone and intonation.

[0344] Step 8:

[0345] The generated audio data is sent from the server to the user's terminal. The terminal converts the received audio data into a playable format and provides it to the user.

[0346] Step 9:

[0347] The device plays audio generated through speakers or earphones to the user. By listening to this personalized audio, the user can feel an emotional connection, as if the deceased person were present.

[0348] (Example 2)

[0349] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0350] In modern society, there is a need for ways to more intimately recreate memories of the deceased and maintain an emotional connection. In particular, providing users with the opportunity to find emotional healing through the voice of the deceased is crucial. However, existing technologies have limitations in adapting the voice of the deceased to their emotions and improving voice quality. Furthermore, the lack of voice generation methods that consider the user's emotional state makes it difficult to provide a personalized conversational experience.

[0351] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0352] In this invention, the server includes means for receiving voice information of the deceased, means for extracting acoustic features from the received voice information, means for reducing noise from the acoustic features, means for evaluating the user's emotional state using an emotion analysis engine, means for reflecting the evaluated emotional state in a generation model and converting the information into speech, and means for distributing the generated speech to the user's terminal. This makes it possible to reproduce the voice of the deceased in high quality and in a manner adapted to the user's emotions, thereby providing an individualized emotional dialogue experience.

[0353] "Voice information of the deceased" refers to audio data of an individual recorded in the past, and this data serves as the basic information for analyzing and reconstructing the voice of the deceased.

[0354] "Acoustic features" are numerical indicators that show the attributes of speech, such as Mel-frequency cepstrum coefficients, pitch, and waveform data extracted from speech data.

[0355] "Methods for reducing noise" refer to processes that remove unwanted noise from audio data and improve sound quality.

[0356] A "generative model" is a model trained using machine learning techniques, and it is an algorithm that generates speech based on text and emotional states.

[0357] An "emotion analysis engine" is software or a system that evaluates a user's emotional state based on the characteristics of input text or audio.

[0358] "User's emotional state" refers to the emotional state a user exhibits when using voice input or text input, and includes emotions such as joy, sadness, and anger.

[0359] "User's device" refers to a device used to receive and play back generated audio data, and includes smartphones, personal computers, and other similar devices.

[0360] This invention provides a more personalized conversational experience in a system that reproduces the voice of a deceased person by taking into account the user's emotional state. Details for implementing this system are provided below.

[0361] 1. Processing of audio data:

[0362] The user uploads the deceased person's voice data to the server via their device. The server uses voice processing software (e.g., Python libraries librosa and pydub) to extract acoustic features. This step involves a noise reduction process to improve the quality of the voice data.

[0363] 2. Emotion analysis and speech synthesis:

[0364] The server uses a generative AI model (e.g., Tacotron 2 or WaveNet) to train a speech synthesis model based on extracted acoustic features. When the user inputs text they want to reproduce in the deceased person's voice from their device, their emotional state is analyzed through an emotion analysis engine (e.g., a natural language processing API). This analysis result is provided to the generative AI model as a prompt along with the text information.

[0365] 3. Speech generation and output:

[0366] The generative AI model generates speech that reflects the user's emotional state based on the user's input text. The generated speech data is sent from the server to the user's device and played back on the device. Through this process, the user can experience an emotional connection with the voice of a deceased loved one.

[0367] Specific example:

[0368] For example, if a user types "Today was a good day," the sentiment analysis engine interprets this as joy and feeds it back to the generative model. Based on this information, the server generates speech in a soft, cheerful tone. An example of a prompt is: "The user's emotion is joy. Please reproduce 'Today was a good day' in the voice of the deceased."

[0369] This embodiment allows users to enjoy an emotionally resonant audio experience. The system aims to provide users with emotional comfort by offering an advanced audio playback method using the voice of a deceased person.

[0370] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0371] Step 1:

[0372] The user uploads the deceased person's voice data to the server via their device. The input voice data file is selected on the device and sent to the server. The user's specific operation involves selecting the voice file on the file selection screen and pressing the "Upload" button. The output is the voice data saved on the server.

[0373] Step 2:

[0374] The server extracts acoustic features from the received audio data. This process uses audio analysis software (e.g., the librosa library in Python) to perform noise reduction and calculate Mel-frequency cepstrum coefficients (MFCCs). The input is the audio data obtained in step 1, and the output is acoustic feature data with reduced noise.

[0375] Step 3:

[0376] The server sends text data entered from the terminal to the sentiment analysis engine in order to analyze the user's emotions. The text entered by the user on the terminal is subject to sentiment analysis. Accordingly, a natural language processing API is used to determine the emotional state of the input text. The output is data indicating the analyzed emotional state.

[0377] Step 4:

[0378] The server uses emotion data and acoustic features obtained from the emotion analysis engine to perform speech synthesis using a generative AI model. At this stage, the output data from steps 2 and 3 is used as input. The speech synthesis model receives this data as prompts and generates speech appropriate to the emotion. The output is the generated speech data.

[0379] Step 5:

[0380] The server sends the generated audio data to the user's terminal. In this process, the audio data obtained in step 4 is used as input and transferred to the terminal. The output is the audio data received by the terminal.

[0381] Step 6:

[0382] The device plays the received audio data to the user. This process utilizes the device's audio player, and the generated audio is played back to the user through speakers or headphones. This allows the user to have an interactive experience through the voice of the deceased.

[0383] (Application Example 2)

[0384] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0385] In modern times, it is a crucial issue for people to preserve memories of deceased loved ones through audio recordings, thereby gaining emotional connection and healing. While conventional technology could reproduce the voice of a deceased person, it struggled to generate personalized audio that responded to the user's emotions. Therefore, there is a need to provide a system that reflects the user's feelings and reproduces the voice of a deceased person in a more natural and emotionally resonant way.

[0386] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0387] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, and an emotion analysis function for analyzing the user's emotional state. This makes it possible to generate and provide to the user the voice of the deceased after making adjustments according to the user's emotions.

[0388] "Audio data" refers to information that digitizes the waveform of sound, and includes the specific characteristics of the deceased person's voice.

[0389] "Speech features" are individual attributes such as pitch, intensity, and texture extracted from speech data, and these features make it possible to analyze speech patterns.

[0390] A "speech synthesis algorithm" is a computational method for generating digital data that sounds like a human voice based on speech characteristics.

[0391] "Emotion analysis function" is a technology that automatically identifies emotions from user input or voice and quantifies the user's emotional state.

[0392] A "user device" is an electronic device owned by the user, which is a terminal for receiving and playing back generated audio data.

[0393] To implement this invention, a specific speech synthesis system is used. First, the server receives the deceased person's voice data from the user. This voice data is processed on the server, first undergoing noise reduction. Next, voice features are extracted. These voice features include the basic pitch, timbre, and prosody of the deceased person's voice.

[0394] The server trains a speech synthesis algorithm using the extracted speech features. This algorithm could, for example, be a Transformer-based speech synthesis model. The trained model can then accept text provided by the user as input and generate speech data in the deceased person's voice.

[0395] The user terminal has the function of receiving and playing back the generated audio data. Furthermore, an emotion analysis function is introduced to analyze the user's emotions. Emotion analysis can be performed from the user's input text, voice intonation, and facial expression data acquired from the camera. This uses software such as the Python library "librosa" and "OpenCV".

[0396] The results of the emotion analysis are fed back to the speech synthesis model and reflected in the generated speech. For example, if a user says "Today was a good day" to their smartphone, that emotion is analyzed as joy, and the server uses this information to generate speech in a cheerful tone.

[0397] An example of a prompt to input into a generative AI model is the instruction, "The user is requesting X, and their emotion is estimated to be Y. Provide advice in the voice of the deceased that is in harmony with the emotion." This allows for a more personalized voice experience that is attentive to the user's emotional state.

[0398] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0399] Step 1:

[0400] The user uses a smartphone or other device to input the text they want to be reproduced in the deceased person's voice. The entered text is sent from the user's device to the server. At this point, the input is text data and is saved as data ready to be sent to the server as output.

[0401] Step 2:

[0402] The server receives audio data of the deceased person, provided in advance by the user. Since this audio data may contain noise, the server applies a noise reduction algorithm to remove unwanted sounds. The input is the audio data of the deceased person, and the output is the audio data with the noise removed.

[0403] Step 3:

[0404] The server extracts speech features from the denoised audio data. These speech features include pitch, tempo, and voice quality. This process uses signal processing techniques to quantify these features. The input is denoised audio data, and the output is the extracted speech features.

[0405] Step 4:

[0406] The server trains a speech synthesis algorithm using the extracted speech features. Deep learning techniques are used to enhance the algorithm so that it can reproduce the deceased person's speech patterns. The input is speech features, and the output is the trained speech synthesis model.

[0407] Step 5:

[0408] In parallel, the user terminal uses input text, audio, and video data obtained from the camera to analyze the user's emotions. Using tools such as Python's "librosa" and "OpenCV," the user's intonation and facial expressions are analyzed, and their emotional state is quantified. The input is text, audio, and video data, and the output is the user's emotional parameters.

[0409] Step 6:

[0410] The server feeds the user's emotion parameters back into the speech synthesis model, which then converts the input text into speech in the deceased person's voice. At this stage, the speech is generated with a tone and tempo that reflects the user's emotions. The input consists of the user's emotion parameters and text, and the output is emotion-harmonized speech data.

[0411] Step 7:

[0412] The generated audio data is sent from the server to the user's terminal, which then plays the audio. Through this output, the user can experience what it's like to converse with the deceased person. The input is the generated audio data, and the output is the audio itself that is played back.

[0413] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0414] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0415] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0416] [Third Embodiment]

[0417] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0418] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0419] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0420] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0421] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0422] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0423] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0424] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0425] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0426] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0427] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0428] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0429] This invention is a speech synthesis system that reproduces the voice of a deceased person, enabling a dialogue experience between the bereaved family and the deceased. The system begins with the user uploading the deceased person's voice data to a server. The server processes this voice data, removing noise and cleaning the audio, and then extracts voice features. These obtained voice features are used to train a speech synthesis model, which then builds a model that mimics the deceased person's speaking style and voice tone.

[0430] The user provides input text from their device, which is then sent to the server. Based on the input text received from the user, the server uses a trained speech synthesis model to generate audio data that reproduces the deceased person's voice. This generated audio data is then sent back to the user's device and delivered to the user through the device in the deceased person's voice.

[0431] As a concrete example, if a user enters the text "Please tell me how your day was today," the server processes the text using a speech synthesis model and generates audio data that mimics the deceased person's voice. This audio data is sent to the terminal, and the user receives the message in the deceased person's voice, experiencing the feeling as if the deceased person is speaking to them in person.

[0432] In this way, the present invention aims to provide a new connection with the deceased and bring healing to the hearts of bereaved families. This system allows users to look back on past memories and once again have the experience of interacting with the deceased through their voice.

[0433] The following describes the processing flow.

[0434] Step 1:

[0435] Users collect audio data of deceased individuals and upload it to a server via their devices. This includes audio files extracted from past recordings and video clips.

[0436] Step 2:

[0437] The server analyzes the received audio data. First, it performs audio preprocessing such as noise reduction and echo reduction to improve sound quality.

[0438] Step 3:

[0439] The server extracts speech features from the pre-processed audio data. This involves a process of calculating features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants.

[0440] Step 4:

[0441] The server uses the extracted speech features to train a speech synthesis model. This allows the model to acquire the ability to mimic the deceased person's speaking style and tone of voice.

[0442] Step 5:

[0443] The user sends the text they want to reproduce in the deceased person's voice as input text to the server via their device.

[0444] Step 6:

[0445] The server passes the received input text to a speech synthesis model, which generates audio data that mimics the deceased person's voice. In this process, the model synthesizes the input text as an audio waveform.

[0446] Step 7:

[0447] The generated audio data is sent from the server to the user's device. The device then converts this audio data into a playable format.

[0448] Step 8:

[0449] The device plays audio data generated through speakers or headphones to the user. The user listens to messages conveyed in the deceased's voice, experiencing a feeling as if they are conversing with the deceased again.

[0450] (Example 1)

[0451] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0452] Currently, many bereaved families wish to feel a connection with their loved ones, but there are limited ways to communicate using the deceased's voice. Existing voice reproduction technologies struggle to realistically reproduce the voice of the deceased, and are insufficient to help bereaved families reminisce about the past. Therefore, there is a growing expectation for technologies that faithfully reproduce the voice of the deceased, providing an experience as if one were conversing with them.

[0453] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0454] In this invention, the server includes means for receiving audio information, means for analyzing the characteristics of the audio from the received audio information, and means for training an information conversion model using the analyzed characteristics of the audio. This allows for the faithful reproduction of the deceased's voice, enabling the user to relive a conversation with the deceased.

[0455] "Audio information" refers to digital or analog information formats used to represent sound, such as audio data or audio signals.

[0456] "Means of receiving" refers to the technology or device used to receive data transmitted from an external source.

[0457] "Means of analysis" refers to techniques or devices used to analyze information and extract specific features or patterns.

[0458] "Features" refer to the characteristics and attributes of the information being analyzed, and are indicators of the data extracted during the analysis process.

[0459] An "information conversion model" is an algorithm or program used to convert input data into another data format.

[0460] "Training methods" refer to the process of improving the performance of a model or algorithm using data to achieve desired results.

[0461] "Inputted character information" refers to information entered as character data, primarily data expressed in text format.

[0462] "Means of conversion" refers to the technology or process of changing data in one format to another.

[0463] "Means of output" refers to technology or equipment for providing processed data to an external party.

[0464] "Means for removing noise" refers to techniques or devices for removing unwanted acoustic components from audio information.

[0465] "User's device" refers to electronic devices and terminals used by the user, including those capable of receiving, displaying, or playing back data.

[0466] This invention relates to a speech synthesis system that reproduces the voice of a deceased person and provides bereaved families with an experience of interacting with the deceased. This system is primarily built through the interaction of a server, a terminal, and the user.

[0467] The user uploads the deceased person's voice data to the server via their device. The server uses an audio editing library (e.g., librosa) to remove noise from the data and improve the sound quality. Then, it uses an audio analysis tool (e.g., Praat) to extract the voice's characteristics. This results in the acquisition of data such as the voice's pitch, range, and timbre.

[0468] Next, the server trains a speech synthesis model. This model uses a machine learning framework (e.g., TensorFlow, PyTorch). The training is performed to mimic the deceased person's unique speaking style and tone of voice based on the extracted speech features. This creates an information transformation model that can naturally reproduce the deceased person's voice.

[0469] The user enters a message they wish to convey to the deceased as text into the terminal. When this text is sent to the server, the server uses a trained model to convert the text into speech. During this process, the prompt "Please mimic everyday conversation in the voice of the deceased" is taken into consideration.

[0470] As a concrete example, if a user enters the text "Please tell me how your day was," the server converts this into audio data. The generated audio data is reproduced in the deceased person's voice and provided to the user through their device. This process allows the user to feel as if they are conversing with the deceased again and to immerse themselves in past memories through that voice.

[0471] This system aims to help bereaved families rebuild their connection with the deceased and find emotional healing.

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

[0473] Step 1:

[0474] The user uploads the deceased person's voice data to the server using their device. The input consists of the deceased person's voice files (e.g., WAV, MP3 format). The server receives these files and prepares for subsequent audio processing.

[0475] Step 2:

[0476] The server removes noise and cleans the received audio data. The input is the uploaded audio data, and background noise and unwanted sounds are removed using a noise reduction library (e.g., librosa). The output is cleaned, high-quality audio data. Specific operations include filtering.

[0477] Step 3:

[0478] The server extracts audio features from the cleaned audio data. The input is processed audio data, and an audio analysis tool (e.g., Praat) is used to extract numerical characteristics such as pitch, range, and intensity. The output is an audio feature vector. This process involves specific frequency analysis and time-series analysis.

[0479] Step 4:

[0480] The server trains a speech synthesis model. The input is a vector of speech features, and a speech synthesis model capable of reproducing the voice of a deceased person is built through a machine learning framework (e.g., TensorFlow, PyTorch). The output is the trained speech synthesis model. Specifically, model parameters are optimized and adjusted through a feedback loop.

[0481] Step 5:

[0482] The user sends input text to the server via their device. The server receives text information specified by the user as input (e.g., "Please tell me how your day was today").

[0483] Step 6:

[0484] The server uses a trained speech synthesis model to convert input text into speech data. The input consists of text from the user and the trained model, and the generative AI model converts the text data into speech data. Here, the prompt "Imitate an everyday conversation in the voice of the deceased" is considered. The output is speech data that imitates the voice of the deceased.

[0485] Step 7:

[0486] The server sends the generated audio data to the user's device. The input is the generated audio data, which the user's device receives. The device then uses an audio player or similar device to play the deceased person's voice for the user. The output is the audio message that reaches the user.

[0487] (Application Example 1)

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

[0489] Maintaining memories and the experience of interacting with a deceased loved one through their voice is a means of spiritual healing for many, but in reality, the means to achieve this are very limited. In particular, dynamically reproducing the voice of a deceased person in response to the surrounding environment or specific situations has been difficult with conventional technology. In response to this, there is a need for technology that allows users to experience the voice of a deceased person more realistically in a virtual environment, thereby allowing them to feel past memories more deeply.

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

[0491] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, means for converting input text into voice data based on a trained speech synthesis model, means for detecting location information for playing the deceased's voice within a virtual environment, and means for dynamically generating and playing the deceased's voice data according to specific circumstances within the virtual environment. This enables the user to have an interactive experience using the voice of the deceased within the virtual environment.

[0492] "Voice data of a deceased person" refers to a collection of audio signals in which the voice of a deceased person has been recorded and saved.

[0493] "Speech features" are data extracted from acoustic signals that represent the nature and phonetic characteristics of a voice.

[0494] A "speech synthesis model" is an algorithm or structure for generating speech data that mimics a human voice based on speech features.

[0495] "Input text" refers to a string of characters that a user manually or automatically enters to give instructions or information to a computer.

[0496] "Audio data" refers to a collection of information recorded from sound and stored in digital or analog format.

[0497] A "virtual environment" is a space or situation artificially constructed using computer technology, providing users with experiences that do not exist in reality.

[0498] "Location information" refers to data that indicates the physical location of a specific object or user, and is usually expressed in coordinate format.

[0499] "Means of dynamically generating and reproducing information" refers to methods and processes for generating and providing information to users in real time, while changing it according to the situation and conditions.

[0500] The system that realizes this invention includes a series of processes to reproduce the voice of a deceased person and allow the user to experience it in a virtual environment. The server receives the deceased person's voice data and processes it to extract voice features. This includes data preprocessing using a noise reduction algorithm to clean up the voice data. Acoustic signal analysis software is used to extract voice features. Subsequently, machine learning libraries such as PyTorch are used to train a speech synthesis model. This model can reproduce the deceased person's voice based on the extracted voice features.

[0501] The server then generates audio data using a speech synthesis model based on the text entered by the user. This generation process is performed using the Google Cloud Text-to-Speech API, and the speech conversion is carried out in real time. The generated audio data is sent to the user's smart glasses or head-mounted display. This allows the user to hear the voice of the deceased based on specific location information while moving around in the virtual environment.

[0502] By utilizing sensors built into the device, the system acquires the user's location information in real time and dynamically plays audio according to the location. For this purpose, Firebase is used for information management, and audio messages associated with the user's location and selected virtual items are provided.

[0503] As a concrete example, when a user reaches a specific store in a virtual shopping mall, a voice recording of the deceased reminiscing about their favorite items is played. An example of a prompt message would be: "Create a message using the deceased's voice, telling the user about Christmas memories, and adding specific anecdotes that evoke warm family moments." This allows the user to have an experience as if they were having a conversation with the deceased.

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

[0505] Step 1:

[0506] The server receives audio data of the deceased from the user. This input data is preprocessed to verify the consistency of its format and sample rate, and then saved as an audio file.

[0507] Step 2:

[0508] The server extracts speech features from the received audio data. Here, speech signal processing techniques are used to extract Mel-frequency cepstrum coefficients (MFCCs), among other things. This data processing is intended to numerically represent the characteristics of the speech data, and the results are output as a list of speech features.

[0509] Step 3:

[0510] The server trains a speech synthesis model using speech features. Using machine learning libraries (e.g., PyTorch), it constructs models such as multilayer perceptrons and recurrent neural networks using the speech features as input, and then adjusts the model parameters. The output is a trained model capable of mimicking the voice of a deceased person.

[0511] Step 4:

[0512] The user enters the input text to be used in the virtual environment into the terminal. This text is sent to the server as the message the user wants the deceased to say.

[0513] Step 5:

[0514] The server uses a trained speech synthesis model based on the received text to generate speech data. The Google Cloud Text-to-Speech API is used to convert the input text into speech data and generate it as data.

[0515] Step 6:

[0516] The server sends the generated audio data to the user's device. The device receives this audio data and prepares to play it back in the deceased person's voice.

[0517] Step 7:

[0518] The user's device uses its built-in sensors to detect its position within the virtual environment in real time. This location information is output as location coordinates because it affects the timing of audio data playback.

[0519] Step 8:

[0520] Based on its location, the device dynamically plays audio data received from the server when it reaches a specific point within the virtual environment. Specifically, a message in the deceased's voice, linked to the location, is played, allowing the user to experience a conversation with the deceased.

[0521] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0522] This invention combines a system for reproducing the voice of a deceased person with an emotion engine that recognizes the user's emotions. The system begins by uploading the deceased person's voice data to a server. The server removes noise from the received voice data and extracts voice features. This results in higher quality voice data, which is used to train a speech synthesis model.

[0523] The user inputs text they wish to be reproduced in the deceased person's voice through their device, while an emotion engine analyzes the user's emotional state. The emotion engine analyzes the input text, voice intonation, volume, tempo, etc., to determine the user's emotions. This analysis result is fed back to the speech synthesis model, influencing the generated voice data. As a result, the generated voice is in harmony with the user's emotions, creating a more personalized conversational experience.

[0524] For example, if a user types "Today was a good day," the emotion engine analyzes the user's emotion as joy and sends that information to the speech synthesis model. Based on this information, the server generates a voice with a softer, more cheerful tone and sends it to the user's device. The device plays the voice, and the user can experience what it's like to share joy with a loved one through their voice.

[0525] Based on this format, users can have an emotional connection with the deceased through recreated voices, allowing them to find solace. Furthermore, the introduction of an emotion engine makes the conversational experience more personalized, providing communication that is attentive to the user's emotions at any given moment.

[0526] The following describes the processing flow.

[0527] Step 1:

[0528] Users collect audio data of deceased individuals and upload it to the server. This includes audio clips and audio files extracted from videos.

[0529] Step 2:

[0530] The server receives the uploaded audio data and removes noise from it. It analyzes the audio to reduce unwanted background noise and performs pre-processing to improve sound quality.

[0531] Step 3:

[0532] From the processed audio data, the server extracts audio features. Specifically, it calculates acoustic features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants to obtain the deceased person's characteristic voice pattern.

[0533] Step 4:

[0534] The server trains a speech synthesis model based on the extracted speech features. This training enables the model to accurately mimic the deceased person's voice quality and speaking style.

[0535] Step 5:

[0536] Users input text they want to reproduce in the deceased's voice via their device and send it to the server. The text they input is what they wanted to exchange with the deceased or any messages they wished they had.

[0537] Step 6:

[0538] The emotion engine operates on the server, analyzing the user's input text and their current voice or input method to recognize their emotional state. This analysis determines whether the user is expressing a specific emotion, such as joy, sadness, or excitement.

[0539] Step 7:

[0540] Based on the analysis results, the server passes the input text to a speech synthesis model and generates audio data by applying emotional feedback. This adjusts the deceased person's voice to match the user's emotions in terms of tone and intonation.

[0541] Step 8:

[0542] The generated audio data is sent from the server to the user's terminal. The terminal converts the received audio data into a playable format and provides it to the user.

[0543] Step 9:

[0544] The device plays audio generated through speakers or earphones to the user. By listening to this personalized audio, the user can feel an emotional connection, as if the deceased person were present.

[0545] (Example 2)

[0546] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0547] In modern society, there is a need for ways to more intimately recreate memories of the deceased and maintain an emotional connection. In particular, providing users with the opportunity to find emotional healing through the voice of the deceased is crucial. However, existing technologies have limitations in adapting the voice of the deceased to their emotions and improving voice quality. Furthermore, the lack of voice generation methods that consider the user's emotional state makes it difficult to provide a personalized conversational experience.

[0548] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0549] In this invention, the server includes means for receiving voice information of the deceased, means for extracting acoustic features from the received voice information, means for reducing noise from the acoustic features, means for evaluating the user's emotional state using an emotion analysis engine, means for reflecting the evaluated emotional state in a generation model and converting the information into speech, and means for distributing the generated speech to the user's terminal. This makes it possible to reproduce the voice of the deceased in high quality and in a manner adapted to the user's emotions, thereby providing an individualized emotional dialogue experience.

[0550] "Voice information of the deceased" refers to audio data of an individual recorded in the past, and this data serves as the basic information for analyzing and reconstructing the voice of the deceased.

[0551] "Acoustic features" are numerical indicators that show the attributes of speech, such as Mel-frequency cepstrum coefficients, pitch, and waveform data extracted from speech data.

[0552] "Methods for reducing noise" refer to processes that remove unwanted noise from audio data and improve sound quality.

[0553] A "generative model" is a model trained using machine learning techniques, and it is an algorithm that generates speech based on text and emotional states.

[0554] An "emotion analysis engine" is software or a system that evaluates a user's emotional state based on the characteristics of input text or audio.

[0555] "User's emotional state" refers to the emotional state a user exhibits when using voice input or text input, and includes emotions such as joy, sadness, and anger.

[0556] "User's device" refers to a device used to receive and play back generated audio data, and includes smartphones, personal computers, and other similar devices.

[0557] This invention provides a more personalized conversational experience in a system that reproduces the voice of a deceased person by taking into account the user's emotional state. Details for implementing this system are provided below.

[0558] 1. Processing of audio data:

[0559] The user uploads the deceased person's voice data to the server via their device. The server uses voice processing software (e.g., Python libraries librosa and pydub) to extract acoustic features. This step involves a noise reduction process to improve the quality of the voice data.

[0560] 2. Emotion analysis and speech synthesis:

[0561] The server uses a generative AI model (e.g., Tacotron 2 or WaveNet) to train a speech synthesis model based on extracted acoustic features. When the user inputs text they want to reproduce in the deceased person's voice from their device, their emotional state is analyzed through an emotion analysis engine (e.g., a natural language processing API). This analysis result is provided to the generative AI model as a prompt along with the text information.

[0562] 3. Speech generation and output:

[0563] The generative AI model generates speech that reflects the user's emotional state based on the user's input text. The generated speech data is sent from the server to the user's device and played back on the device. Through this process, the user can experience an emotional connection with the voice of a deceased loved one.

[0564] Specific example:

[0565] For example, if a user types "Today was a good day," the sentiment analysis engine interprets this as joy and feeds it back to the generative model. Based on this information, the server generates speech in a soft, cheerful tone. An example of a prompt is: "The user's emotion is joy. Please reproduce 'Today was a good day' in the voice of the deceased."

[0566] This embodiment allows users to enjoy an emotionally resonant audio experience. The system aims to provide users with emotional comfort by offering an advanced audio playback method using the voice of a deceased person.

[0567] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0568] Step 1:

[0569] The user uploads the deceased person's voice data to the server via their device. The input voice data file is selected on the device and sent to the server. The user's specific operation involves selecting the voice file on the file selection screen and pressing the "Upload" button. The output is the voice data saved on the server.

[0570] Step 2:

[0571] The server extracts acoustic features from the received audio data. This process uses audio analysis software (e.g., the librosa library in Python) to perform noise reduction and calculate Mel-frequency cepstrum coefficients (MFCCs). The input is the audio data obtained in step 1, and the output is acoustic feature data with reduced noise.

[0572] Step 3:

[0573] The server sends text data entered from the terminal to the sentiment analysis engine in order to analyze the user's emotions. The text entered by the user on the terminal is subject to sentiment analysis. Accordingly, a natural language processing API is used to determine the emotional state of the input text. The output is data indicating the analyzed emotional state.

[0574] Step 4:

[0575] The server uses emotion data and acoustic features obtained from the emotion analysis engine to perform speech synthesis using a generative AI model. At this stage, the output data from steps 2 and 3 is used as input. The speech synthesis model receives this data as prompts and generates speech appropriate to the emotion. The output is the generated speech data.

[0576] Step 5:

[0577] The server sends the generated audio data to the user's terminal. In this process, the audio data obtained in step 4 is used as input and transferred to the terminal. The output is the audio data received by the terminal.

[0578] Step 6:

[0579] The device plays the received audio data to the user. This process utilizes the device's audio player, and the generated audio is played back to the user through speakers or headphones. This allows the user to have an interactive experience through the voice of the deceased.

[0580] (Application Example 2)

[0581] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the headset-type terminal 314 will be referred to as the "terminal."

[0582] In modern times, it is a crucial issue for people to preserve memories of deceased loved ones through audio recordings, thereby gaining emotional connection and healing. While conventional technology could reproduce the voice of a deceased person, it struggled to generate personalized audio that responded to the user's emotions. Therefore, there is a need to provide a system that reflects the user's feelings and reproduces the voice of a deceased person in a more natural and emotionally resonant way.

[0583] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0584] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, and an emotion analysis function for analyzing the user's emotional state. This makes it possible to generate and provide to the user the voice of the deceased after making adjustments according to the user's emotions.

[0585] "Audio data" refers to information that digitizes the waveform of sound, and includes the specific characteristics of the deceased person's voice.

[0586] "Speech features" are individual attributes such as pitch, intensity, and texture extracted from speech data, and these features make it possible to analyze speech patterns.

[0587] A "speech synthesis algorithm" is a computational method for generating digital data that sounds like a human voice based on speech characteristics.

[0588] "Emotion analysis function" is a technology that automatically identifies emotions from user input or voice and quantifies the user's emotional state.

[0589] A "user device" is an electronic device owned by the user, which is a terminal for receiving and playing back generated audio data.

[0590] To implement this invention, a specific speech synthesis system is used. First, the server receives the deceased person's voice data from the user. This voice data is processed on the server, first undergoing noise reduction. Next, voice features are extracted. These voice features include the basic pitch, timbre, and prosody of the deceased person's voice.

[0591] The server trains a speech synthesis algorithm using the extracted speech features. This algorithm could, for example, be a Transformer-based speech synthesis model. The trained model can then accept text provided by the user as input and generate speech data in the deceased person's voice.

[0592] The user terminal has the function of receiving and playing back the generated audio data. Furthermore, an emotion analysis function is introduced to analyze the user's emotions. Emotion analysis can be performed from the user's input text, voice intonation, and facial expression data acquired from the camera. This uses software such as the Python library "librosa" and "OpenCV".

[0593] The results of the emotion analysis are fed back to the speech synthesis model and reflected in the generated speech. For example, if a user says "Today was a good day" to their smartphone, that emotion is analyzed as joy, and the server uses this information to generate speech in a cheerful tone.

[0594] An example of a prompt to input into a generative AI model is the instruction, "The user is requesting X, and their emotion is estimated to be Y. Provide advice in the voice of the deceased that is in harmony with the emotion." This allows for a more personalized voice experience that is attentive to the user's emotional state.

[0595] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0596] Step 1:

[0597] The user uses a smartphone or other device to input the text they want to be reproduced in the deceased person's voice. The entered text is sent from the user's device to the server. At this point, the input is text data and is saved as data ready to be sent to the server as output.

[0598] Step 2:

[0599] The server receives audio data of the deceased person, provided in advance by the user. Since this audio data may contain noise, the server applies a noise reduction algorithm to remove unwanted sounds. The input is the audio data of the deceased person, and the output is the audio data with the noise removed.

[0600] Step 3:

[0601] The server extracts speech features from the denoised audio data. These speech features include pitch, tempo, and voice quality. This process uses signal processing techniques to quantify these features. The input is denoised audio data, and the output is the extracted speech features.

[0602] Step 4:

[0603] The server trains a speech synthesis algorithm using the extracted speech features. Deep learning techniques are used to enhance the algorithm so that it can reproduce the deceased person's speech patterns. The input is speech features, and the output is the trained speech synthesis model.

[0604] Step 5:

[0605] In parallel, the user terminal uses input text, audio, and video data obtained from the camera to analyze the user's emotions. Using tools such as Python's "librosa" and "OpenCV," the user's intonation and facial expressions are analyzed, and their emotional state is quantified. The input is text, audio, and video data, and the output is the user's emotional parameters.

[0606] Step 6:

[0607] The server feeds the user's emotion parameters back into the speech synthesis model, which then converts the input text into speech in the deceased person's voice. At this stage, the speech is generated with a tone and tempo that reflects the user's emotions. The input consists of the user's emotion parameters and text, and the output is emotion-harmonized speech data.

[0608] Step 7:

[0609] The generated audio data is sent from the server to the user's terminal, which then plays the audio. Through this output, the user can experience what it's like to converse with the deceased person. The input is the generated audio data, and the output is the audio itself that is played back.

[0610] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0611] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0612] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[0613] [Fourth Embodiment]

[0614] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0615] As shown in Figure 7, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0616] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0617] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[0618] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0619] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0620] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0621] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0622] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0623] The specific processing program 56 is an example of a "program" relating to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0624] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0625] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0626] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0627] This invention is a speech synthesis system that reproduces the voice of a deceased person, enabling a dialogue experience between the bereaved family and the deceased. The system begins with the user uploading the deceased person's voice data to a server. The server processes this voice data, removing noise and cleaning the audio, and then extracts voice features. These obtained voice features are used to train a speech synthesis model, which then builds a model that mimics the deceased person's speaking style and voice tone.

[0628] The user provides input text from their device, which is then sent to the server. Based on the input text received from the user, the server uses a trained speech synthesis model to generate audio data that reproduces the deceased person's voice. This generated audio data is then sent back to the user's device and delivered to the user through the device in the deceased person's voice.

[0629] As a concrete example, if a user enters the text "Please tell me how your day was today," the server processes the text using a speech synthesis model and generates audio data that mimics the deceased person's voice. This audio data is sent to the terminal, and the user receives the message in the deceased person's voice, experiencing the feeling as if the deceased person is speaking to them in person.

[0630] In this way, the present invention aims to provide a new connection with the deceased and bring healing to the hearts of bereaved families. This system allows users to look back on past memories and once again have the experience of interacting with the deceased through their voice.

[0631] The following describes the processing flow.

[0632] Step 1:

[0633] Users collect audio data of deceased individuals and upload it to a server via their devices. This includes audio files extracted from past recordings and video clips.

[0634] Step 2:

[0635] The server analyzes the received audio data. First, it performs audio preprocessing such as noise reduction and echo reduction to improve sound quality.

[0636] Step 3:

[0637] The server extracts speech features from the pre-processed audio data. This involves a process of calculating features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants.

[0638] Step 4:

[0639] The server uses the extracted speech features to train a speech synthesis model. This allows the model to acquire the ability to mimic the deceased person's speaking style and tone of voice.

[0640] Step 5:

[0641] The user sends the text they want to reproduce in the deceased person's voice as input text to the server via their device.

[0642] Step 6:

[0643] The server passes the received input text to a speech synthesis model, which generates audio data that mimics the deceased person's voice. In this process, the model synthesizes the input text as an audio waveform.

[0644] Step 7:

[0645] The generated audio data is sent from the server to the user's device. The device then converts this audio data into a playable format.

[0646] Step 8:

[0647] The device plays audio data generated through speakers or headphones to the user. The user listens to messages conveyed in the deceased's voice, experiencing a feeling as if they are conversing with the deceased again.

[0648] (Example 1)

[0649] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0650] Currently, many bereaved families wish to feel a connection with their loved ones, but there are limited ways to communicate using the deceased's voice. Existing voice reproduction technologies struggle to realistically reproduce the voice of the deceased, and are insufficient to help bereaved families reminisce about the past. Therefore, there is a growing expectation for technologies that faithfully reproduce the voice of the deceased, providing an experience as if one were conversing with them.

[0651] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0652] In this invention, the server includes means for receiving audio information, means for analyzing the characteristics of the audio from the received audio information, and means for training an information conversion model using the analyzed characteristics of the audio. This allows for the faithful reproduction of the deceased's voice, enabling the user to relive a conversation with the deceased.

[0653] "Audio information" refers to digital or analog information formats used to represent sound, such as audio data or audio signals.

[0654] "Means of receiving" refers to the technology or device used to receive data transmitted from an external source.

[0655] "Means of analysis" refers to techniques or devices used to analyze information and extract specific features or patterns.

[0656] "Features" refer to the characteristics and attributes of the information being analyzed, and are indicators of the data extracted during the analysis process.

[0657] An "information conversion model" is an algorithm or program used to convert input data into another data format.

[0658] "Training methods" refer to the process of improving the performance of a model or algorithm using data to achieve desired results.

[0659] "Inputted character information" refers to information entered as character data, primarily data expressed in text format.

[0660] "Means of conversion" refers to the technology or process of changing data in one format to another.

[0661] "Means of output" refers to technology or equipment for providing processed data to an external party.

[0662] "Means for removing noise" refers to techniques or devices for removing unwanted acoustic components from audio information.

[0663] "User's device" refers to electronic devices and terminals used by the user, including those capable of receiving, displaying, or playing back data.

[0664] This invention relates to a speech synthesis system that reproduces the voice of a deceased person and provides bereaved families with an experience of interacting with the deceased. This system is primarily built through the interaction of a server, a terminal, and the user.

[0665] The user uploads the deceased person's voice data to the server via their device. The server uses an audio editing library (e.g., librosa) to remove noise from the data and improve the sound quality. Then, it uses an audio analysis tool (e.g., Praat) to extract the voice's characteristics. This results in the acquisition of data such as the voice's pitch, range, and timbre.

[0666] Next, the server trains a speech synthesis model. This model uses a machine learning framework (e.g., TensorFlow, PyTorch). The training is performed to mimic the deceased person's unique speaking style and tone of voice based on the extracted speech features. This creates an information transformation model that can naturally reproduce the deceased person's voice.

[0667] The user enters a message they wish to convey to the deceased as text into the terminal. When this text is sent to the server, the server uses a trained model to convert the text into speech. During this process, the prompt "Please mimic everyday conversation in the voice of the deceased" is taken into consideration.

[0668] As a concrete example, if a user enters the text "Please tell me how your day was," the server converts this into audio data. The generated audio data is reproduced in the deceased person's voice and provided to the user through their device. This process allows the user to feel as if they are conversing with the deceased again and to immerse themselves in past memories through that voice.

[0669] This system aims to help bereaved families rebuild their connection with the deceased and find emotional healing.

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

[0671] Step 1:

[0672] The user uploads the deceased person's voice data to the server using their device. The input consists of the deceased person's voice files (e.g., WAV, MP3 format). The server receives these files and prepares for subsequent audio processing.

[0673] Step 2:

[0674] The server removes noise and cleans the received audio data. The input is the uploaded audio data, and background noise and unwanted sounds are removed using a noise reduction library (e.g., librosa). The output is cleaned, high-quality audio data. Specific operations include filtering.

[0675] Step 3:

[0676] The server extracts audio features from the cleaned audio data. The input is processed audio data, and an audio analysis tool (e.g., Praat) is used to extract numerical characteristics such as pitch, range, and intensity. The output is an audio feature vector. This process involves specific frequency analysis and time-series analysis.

[0677] Step 4:

[0678] The server trains a speech synthesis model. The input is a vector of speech features, and a speech synthesis model capable of reproducing the voice of a deceased person is built through a machine learning framework (e.g., TensorFlow, PyTorch). The output is the trained speech synthesis model. Specifically, model parameters are optimized and adjusted through a feedback loop.

[0679] Step 5:

[0680] The user sends input text to the server via their device. The server receives text information specified by the user as input (e.g., "Please tell me how your day was today").

[0681] Step 6:

[0682] The server uses a trained speech synthesis model to convert input text into speech data. The input consists of text from the user and the trained model, and the generative AI model converts the text data into speech data. Here, the prompt "Imitate an everyday conversation in the voice of the deceased" is considered. The output is speech data that imitates the voice of the deceased.

[0683] Step 7:

[0684] The server sends the generated audio data to the user's device. The input is the generated audio data, which the user's device receives. The device then uses an audio player or similar device to play the deceased person's voice for the user. The output is the audio message that reaches the user.

[0685] (Application Example 1)

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

[0687] Maintaining memories and the experience of interacting with a deceased loved one through their voice is a means of spiritual healing for many, but in reality, the means to achieve this are very limited. In particular, dynamically reproducing the voice of a deceased person in response to the surrounding environment or specific situations has been difficult with conventional technology. In response to this, there is a need for technology that allows users to experience the voice of a deceased person more realistically in a virtual environment, thereby allowing them to feel past memories more deeply.

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

[0689] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, means for converting input text into voice data based on a trained speech synthesis model, means for detecting location information for playing the deceased's voice within a virtual environment, and means for dynamically generating and playing the deceased's voice data according to specific circumstances within the virtual environment. This enables the user to have an interactive experience using the voice of the deceased within the virtual environment.

[0690] "Voice data of a deceased person" refers to a collection of audio signals in which the voice of a deceased person has been recorded and saved.

[0691] "Speech features" are data extracted from acoustic signals that represent the nature and phonetic characteristics of a voice.

[0692] A "speech synthesis model" is an algorithm or structure for generating speech data that mimics a human voice based on speech features.

[0693] "Input text" refers to a string of characters that a user manually or automatically enters to give instructions or information to a computer.

[0694] "Audio data" refers to a collection of information recorded from sound and stored in digital or analog format.

[0695] A "virtual environment" is a space or situation artificially constructed using computer technology, providing users with experiences that do not exist in reality.

[0696] "Location information" refers to data that indicates the physical location of a specific object or user, and is usually expressed in coordinate format.

[0697] "Means of dynamically generating and reproducing information" refers to methods and processes for generating and providing information to users in real time, while changing it according to the situation and conditions.

[0698] The system that realizes this invention includes a series of processes to reproduce the voice of a deceased person and allow the user to experience it in a virtual environment. The server receives the deceased person's voice data and processes it to extract voice features. This includes data preprocessing using a noise reduction algorithm to clean up the voice data. Acoustic signal analysis software is used to extract voice features. Subsequently, machine learning libraries such as PyTorch are used to train a speech synthesis model. This model can reproduce the deceased person's voice based on the extracted voice features.

[0699] The server then generates audio data using a speech synthesis model based on the text entered by the user. This generation process is performed using the Google Cloud Text-to-Speech API, and the speech conversion is carried out in real time. The generated audio data is sent to the user's smart glasses or head-mounted display. This allows the user to hear the voice of the deceased based on specific location information while moving around in the virtual environment.

[0700] By utilizing sensors built into the device, the system acquires the user's location information in real time and dynamically plays audio according to the location. For this purpose, Firebase is used for information management, and audio messages associated with the user's location and selected virtual items are provided.

[0701] As a concrete example, when a user reaches a specific store in a virtual shopping mall, a voice recording of the deceased reminiscing about their favorite items is played. An example of a prompt message would be: "Create a message using the deceased's voice, telling the user about Christmas memories, and adding specific anecdotes that evoke warm family moments." This allows the user to have an experience as if they were having a conversation with the deceased.

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

[0703] Step 1:

[0704] The server receives audio data of the deceased from the user. This input data is preprocessed to verify the consistency of its format and sample rate, and then saved as an audio file.

[0705] Step 2:

[0706] The server extracts speech features from the received audio data. Here, speech signal processing techniques are used to extract Mel-frequency cepstrum coefficients (MFCCs), among other things. This data processing is intended to numerically represent the characteristics of the speech data, and the results are output as a list of speech features.

[0707] Step 3:

[0708] The server trains a speech synthesis model using speech features. Using machine learning libraries (e.g., PyTorch), it constructs models such as multilayer perceptrons and recurrent neural networks using the speech features as input, and then adjusts the model parameters. The output is a trained model capable of mimicking the voice of a deceased person.

[0709] Step 4:

[0710] The user enters the input text to be used in the virtual environment into the terminal. This text is sent to the server as the message the user wants the deceased to say.

[0711] Step 5:

[0712] The server uses a trained speech synthesis model based on the received text to generate speech data. The Google Cloud Text-to-Speech API is used to convert the input text into speech data and generate it as data.

[0713] Step 6:

[0714] The server sends the generated audio data to the user's device. The device receives this audio data and prepares to play it back in the deceased person's voice.

[0715] Step 7:

[0716] The user's device uses its built-in sensors to detect its position within the virtual environment in real time. This location information is output as location coordinates because it affects the timing of audio data playback.

[0717] Step 8:

[0718] Based on its location, the device dynamically plays audio data received from the server when it reaches a specific point within the virtual environment. Specifically, a message in the deceased's voice, linked to the location, is played, allowing the user to experience a conversation with the deceased.

[0719] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0720] This invention combines a system for reproducing the voice of a deceased person with an emotion engine that recognizes the user's emotions. The system begins by uploading the deceased person's voice data to a server. The server removes noise from the received voice data and extracts voice features. This results in higher quality voice data, which is used to train a speech synthesis model.

[0721] The user inputs text they wish to be reproduced in the deceased person's voice through their device, while an emotion engine analyzes the user's emotional state. The emotion engine analyzes the input text, voice intonation, volume, tempo, etc., to determine the user's emotions. This analysis result is fed back to the speech synthesis model, influencing the generated voice data. As a result, the generated voice is in harmony with the user's emotions, creating a more personalized conversational experience.

[0722] For example, if a user types "Today was a good day," the emotion engine analyzes the user's emotion as joy and sends that information to the speech synthesis model. Based on this information, the server generates a voice with a softer, more cheerful tone and sends it to the user's device. The device plays the voice, and the user can experience what it's like to share joy with a loved one through their voice.

[0723] Based on this format, users can have an emotional connection with the deceased through recreated voices, allowing them to find solace. Furthermore, the introduction of an emotion engine makes the conversational experience more personalized, providing communication that is attentive to the user's emotions at any given moment.

[0724] The following describes the processing flow.

[0725] Step 1:

[0726] Users collect audio data of deceased individuals and upload it to the server. This includes audio clips and audio files extracted from videos.

[0727] Step 2:

[0728] The server receives the uploaded audio data and removes noise from it. It analyzes the audio to reduce unwanted background noise and performs pre-processing to improve sound quality.

[0729] Step 3:

[0730] From the processed audio data, the server extracts audio features. Specifically, it calculates acoustic features such as Mel-frequency cepstrum coefficients (MFCCs), pitch, and formants to obtain the deceased person's characteristic voice pattern.

[0731] Step 4:

[0732] The server trains a speech synthesis model based on the extracted speech features. This training enables the model to accurately mimic the deceased person's voice quality and speaking style.

[0733] Step 5:

[0734] Users input text they want to reproduce in the deceased's voice via their device and send it to the server. The text they input is what they wanted to exchange with the deceased or any messages they wished they had.

[0735] Step 6:

[0736] The emotion engine operates on the server, analyzing the user's input text and their current voice or input method to recognize their emotional state. This analysis determines whether the user is expressing a specific emotion, such as joy, sadness, or excitement.

[0737] Step 7:

[0738] Based on the analysis results, the server passes the input text to a speech synthesis model and generates audio data by applying emotional feedback. This adjusts the deceased person's voice to match the user's emotions in terms of tone and intonation.

[0739] Step 8:

[0740] The generated audio data is sent from the server to the user's terminal. The terminal converts the received audio data into a playable format and provides it to the user.

[0741] Step 9:

[0742] The device plays audio generated through speakers or earphones to the user. By listening to this personalized audio, the user can feel an emotional connection, as if the deceased person were present.

[0743] (Example 2)

[0744] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0745] In modern society, there is a need for ways to more intimately recreate memories of the deceased and maintain an emotional connection. In particular, providing users with the opportunity to find emotional healing through the voice of the deceased is crucial. However, existing technologies have limitations in adapting the voice of the deceased to their emotions and improving voice quality. Furthermore, the lack of voice generation methods that consider the user's emotional state makes it difficult to provide a personalized conversational experience.

[0746] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0747] In this invention, the server includes means for receiving voice information of the deceased, means for extracting acoustic features from the received voice information, means for reducing noise from the acoustic features, means for evaluating the user's emotional state using an emotion analysis engine, means for reflecting the evaluated emotional state in a generation model and converting the information into speech, and means for distributing the generated speech to the user's terminal. This makes it possible to reproduce the voice of the deceased in high quality and in a manner adapted to the user's emotions, thereby providing an individualized emotional dialogue experience.

[0748] "Voice information of the deceased" refers to audio data of an individual recorded in the past, and this data serves as the basic information for analyzing and reconstructing the voice of the deceased.

[0749] "Acoustic features" are numerical indicators that show the attributes of speech, such as Mel-frequency cepstrum coefficients, pitch, and waveform data extracted from speech data.

[0750] "Methods for reducing noise" refer to processes that remove unwanted noise from audio data and improve sound quality.

[0751] A "generative model" is a model trained using machine learning techniques, and it is an algorithm that generates speech based on text and emotional states.

[0752] An "emotion analysis engine" is software or a system that evaluates a user's emotional state based on the characteristics of input text or audio.

[0753] "User's emotional state" refers to the emotional state a user exhibits when using voice input or text input, and includes emotions such as joy, sadness, and anger.

[0754] "User's device" refers to a device used to receive and play back generated audio data, and includes smartphones, personal computers, and other similar devices.

[0755] This invention provides a more personalized conversational experience in a system that reproduces the voice of a deceased person by taking into account the user's emotional state. Details for implementing this system are provided below.

[0756] 1. Processing of audio data:

[0757] The user uploads the deceased person's voice data to the server via their device. The server uses voice processing software (e.g., Python libraries librosa and pydub) to extract acoustic features. This step involves a noise reduction process to improve the quality of the voice data.

[0758] 2. Emotion analysis and speech synthesis:

[0759] The server uses a generative AI model (e.g., Tacotron 2 or WaveNet) to train a speech synthesis model based on extracted acoustic features. When the user inputs text they want to reproduce in the deceased person's voice from their device, their emotional state is analyzed through an emotion analysis engine (e.g., a natural language processing API). This analysis result is provided to the generative AI model as a prompt along with the text information.

[0760] 3. Speech generation and output:

[0761] The generative AI model generates speech that reflects the user's emotional state based on the user's input text. The generated speech data is sent from the server to the user's device and played back on the device. Through this process, the user can experience an emotional connection with the voice of a deceased loved one.

[0762] Specific example:

[0763] For example, if a user types "Today was a good day," the sentiment analysis engine interprets this as joy and feeds it back to the generative model. Based on this information, the server generates speech in a soft, cheerful tone. An example of a prompt is: "The user's emotion is joy. Please reproduce 'Today was a good day' in the voice of the deceased."

[0764] This embodiment allows users to enjoy an emotionally resonant audio experience. The system aims to provide users with emotional comfort by offering an advanced audio playback method using the voice of a deceased person.

[0765] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0766] Step 1:

[0767] The user uploads the deceased person's voice data to the server via their device. The input voice data file is selected on the device and sent to the server. The user's specific operation involves selecting the voice file on the file selection screen and pressing the "Upload" button. The output is the voice data saved on the server.

[0768] Step 2:

[0769] The server extracts acoustic features from the received audio data. This process uses audio analysis software (e.g., the librosa library in Python) to perform noise reduction and calculate Mel-frequency cepstrum coefficients (MFCCs). The input is the audio data obtained in step 1, and the output is acoustic feature data with reduced noise.

[0770] Step 3:

[0771] The server sends text data entered from the terminal to the sentiment analysis engine in order to analyze the user's emotions. The text entered by the user on the terminal is subject to sentiment analysis. Accordingly, a natural language processing API is used to determine the emotional state of the input text. The output is data indicating the analyzed emotional state.

[0772] Step 4:

[0773] The server uses emotion data and acoustic features obtained from the emotion analysis engine to perform speech synthesis using a generative AI model. At this stage, the output data from steps 2 and 3 is used as input. The speech synthesis model receives this data as prompts and generates speech appropriate to the emotion. The output is the generated speech data.

[0774] Step 5:

[0775] The server sends the generated audio data to the user's terminal. In this process, the audio data obtained in step 4 is used as input and transferred to the terminal. The output is the audio data received by the terminal.

[0776] Step 6:

[0777] The device plays the received audio data to the user. This process utilizes the device's audio player, and the generated audio is played back to the user through speakers or headphones. This allows the user to have an interactive experience through the voice of the deceased.

[0778] (Application Example 2)

[0779] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[0780] In modern times, it is a crucial issue for people to preserve memories of deceased loved ones through audio recordings, thereby gaining emotional connection and healing. While conventional technology could reproduce the voice of a deceased person, it struggled to generate personalized audio that responded to the user's emotions. Therefore, there is a need to provide a system that reflects the user's feelings and reproduces the voice of a deceased person in a more natural and emotionally resonant way.

[0781] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means.

[0782] In this invention, the server includes means for receiving voice data of the deceased, means for extracting voice features from the received voice data, and an emotion analysis function for analyzing the user's emotional state. This makes it possible to generate and provide to the user the voice of the deceased after making adjustments according to the user's emotions.

[0783] "Audio data" refers to information that digitizes the waveform of sound, and includes the specific characteristics of the deceased person's voice.

[0784] "Speech features" are individual attributes such as pitch, intensity, and texture extracted from speech data, and these features make it possible to analyze speech patterns.

[0785] A "speech synthesis algorithm" is a computational method for generating digital data that sounds like a human voice based on speech characteristics.

[0786] "Emotion analysis function" is a technology that automatically identifies emotions from user input or voice and quantifies the user's emotional state.

[0787] A "user device" is an electronic device owned by the user, which is a terminal for receiving and playing back generated audio data.

[0788] To implement this invention, a specific speech synthesis system is used. First, the server receives the deceased person's voice data from the user. This voice data is processed on the server, first undergoing noise reduction. Next, voice features are extracted. These voice features include the basic pitch, timbre, and prosody of the deceased person's voice.

[0789] The server trains a speech synthesis algorithm using the extracted speech features. This algorithm could, for example, be a Transformer-based speech synthesis model. The trained model can then accept text provided by the user as input and generate speech data in the deceased person's voice.

[0790] The user terminal has the function of receiving and playing back the generated audio data. Furthermore, an emotion analysis function is introduced to analyze the user's emotions. Emotion analysis can be performed from the user's input text, voice intonation, and facial expression data acquired from the camera. This uses software such as the Python library "librosa" and "OpenCV".

[0791] The results of the emotion analysis are fed back to the speech synthesis model and reflected in the generated speech. For example, if a user says "Today was a good day" to their smartphone, that emotion is analyzed as joy, and the server uses this information to generate speech in a cheerful tone.

[0792] An example of a prompt to input into a generative AI model is the instruction, "The user is requesting X, and their emotion is estimated to be Y. Provide advice in the voice of the deceased that is in harmony with the emotion." This allows for a more personalized voice experience that is attentive to the user's emotional state.

[0793] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0794] Step 1:

[0795] The user uses a smartphone or other device to input the text they want to be reproduced in the deceased person's voice. The entered text is sent from the user's device to the server. At this point, the input is text data and is saved as data ready to be sent to the server as output.

[0796] Step 2:

[0797] The server receives audio data of the deceased person, provided in advance by the user. Since this audio data may contain noise, the server applies a noise reduction algorithm to remove unwanted sounds. The input is the audio data of the deceased person, and the output is the audio data with the noise removed.

[0798] Step 3:

[0799] The server extracts speech features from the denoised audio data. These speech features include pitch, tempo, and voice quality. This process uses signal processing techniques to quantify these features. The input is denoised audio data, and the output is the extracted speech features.

[0800] Step 4:

[0801] The server trains a speech synthesis algorithm using the extracted speech features. Deep learning techniques are used to enhance the algorithm so that it can reproduce the deceased person's speech patterns. The input is speech features, and the output is the trained speech synthesis model.

[0802] Step 5:

[0803] In parallel, the user terminal uses input text, audio, and video data obtained from the camera to analyze the user's emotions. Using tools such as Python's "librosa" and "OpenCV," the user's intonation and facial expressions are analyzed, and their emotional state is quantified. The input is text, audio, and video data, and the output is the user's emotional parameters.

[0804] Step 6:

[0805] The server feeds the user's emotion parameters back into the speech synthesis model, which then converts the input text into speech in the deceased person's voice. At this stage, the speech is generated with a tone and tempo that reflects the user's emotions. The input consists of the user's emotion parameters and text, and the output is emotion-harmonized speech data.

[0806] Step 7:

[0807] The generated audio data is sent from the server to the user's terminal, which then plays the audio. Through this output, the user can experience what it's like to converse with the deceased person. The input is the generated audio data, and the output is the audio itself that is played back.

[0808] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0809] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0810] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[0811] Furthermore, the emotion identification model 59, acting as an emotion engine, may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0812] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[0813] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[0814] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[0815] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[0816] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[0817] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[0818] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[0819] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[0820] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0821] 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.

[0822] Furthermore, it is not necessary to store the entirety of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[0823] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[0824] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[0825] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[0826] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[0827] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[0828] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted as being incorporated by reference.

[0829] The following is further disclosed regarding the embodiments described above.

[0830] (Claim 1)

[0831] A means of receiving audio data of the deceased,

[0832] A method for extracting speech features from received speech data,

[0833] A means for training a speech synthesis model using extracted speech features,

[0834] A means for converting input text into speech data based on a trained speech synthesis model,

[0835] A means for outputting the generated audio data,

[0836] A system that includes this.

[0837] (Claim 2)

[0838] The system according to claim 1, further comprising means for removing noise from the voice data of a deceased person.

[0839] (Claim 3)

[0840] The system according to claim 1, further comprising means for transmitting audio data generated based on input text to a user terminal.

[0841] "Example 1"

[0842] (Claim 1)

[0843] A means for receiving audio information,

[0844] A means for analyzing the characteristics of speech from received speech information,

[0845] A means of training an information transformation model using the characteristics of the analyzed speech,

[0846] A means for converting input text information into speech information based on a trained information conversion model,

[0847] A means for outputting the converted audio information,

[0848] A system that includes this.

[0849] (Claim 2)

[0850] The system according to claim 1, further comprising means for removing noise from audio information.

[0851] (Claim 3)

[0852] The system according to claim 1, further comprising means for transmitting generated audio information to a user's device.

[0853] "Application Example 1"

[0854] (Claim 1)

[0855] A means of receiving audio data of the deceased,

[0856] A method for extracting speech features from received speech data,

[0857] A means for training a speech synthesis model using extracted speech features,

[0858] A means for converting input text into speech data based on a trained speech synthesis model,

[0859] A means for outputting the generated audio data,

[0860] A means for detecting location information to play back the voice of a deceased person within a virtual environment,

[0861] A means for dynamically generating and playing back the deceased person's voice data according to specific circumstances within a virtual environment,

[0862] A system that includes this.

[0863] (Claim 2)

[0864] The system according to claim 1, further comprising means for removing noise from the voice data of a deceased person.

[0865] (Claim 3)

[0866] The system according to claim 1, further comprising means for transmitting audio data generated based on input text to a user's information terminal.

[0867] "Example 2 of combining an emotion engine"

[0868] (Claim 1)

[0869] Means of receiving voice information of the deceased,

[0870] A method for extracting acoustic features from received voice information,

[0871] Methods for reducing noise based on acoustic characteristics,

[0872] A means of training a generative model using extracted acoustic features,

[0873] A means of evaluating a user's emotional state using an emotion analysis engine,

[0874] A means of converting information into speech by reflecting the evaluated emotional state in a generative model,

[0875] A means of outputting the converted audio,

[0876] A system that includes this.

[0877] (Claim 2)

[0878] The system according to claim 1, further comprising means for delivering audio generated based on input information to the user's terminal.

[0879] (Claim 3)

[0880] The system according to claim 1, further comprising means for removing noise from acoustic characteristics.

[0881] "Application example 2 when combining with an emotional engine"

[0882] (Claim 1)

[0883] A means of receiving audio data of the deceased,

[0884] A method for extracting speech features from received speech data,

[0885] A means for training a speech synthesis algorithm using extracted speech features,

[0886] A means for converting input text into speech data based on a trained speech synthesis algorithm,

[0887] A sentiment analysis function that analyzes the user's emotional state,

[0888] A means of adjusting audio data based on the analyzed emotional state,

[0889] A means for outputting the generated audio data,

[0890] A system that includes this.

[0891] (Claim 2)

[0892] The system according to claim 1, further comprising means for removing unwanted sounds from the voice data of a deceased person.

[0893] (Claim 3)

[0894] The system according to claim 1, further comprising means for transmitting audio data generated based on input text to a user device. [Explanation of Symbols]

[0895] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of receiving audio data of the deceased, A method for extracting speech features from received speech data, A means for training a speech synthesis model using extracted speech features, A means for converting input text into speech data based on a trained speech synthesis model, A means for outputting the generated audio data, A system that includes this.

2. The system according to claim 1, further comprising means for removing noise from the voice data of a deceased person.

3. The system according to claim 1, further comprising means for transmitting audio data generated based on input text to a user terminal.

Images