system

Abstract

We provide the system. [Solution] Voice input method, A pre-processing means for removing noise from the acquired audio signal, A feature extraction means for extracting features from a pre-processed audio signal, A determination means for determining whether the speech was synthesized based on the extracted features, A display means that displays a warning based on the judgment result, A system that includes this.

Description

Technical Field

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

Background Art

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

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In recent years, with the evolution of voice synthesis technology using generative models, realistic synthetic voices can be easily created. While such technology has beneficial uses, the risk of abuse, especially fraud targeting the elderly and the spread of false information, has been increasing. Conventional voice discrimination technology has limitations in accuracy and real-time performance, and there is a need for a system that can accurately and immediately identify synthetic voices.

Means for Solving the Problems

[0005] To solve the above problems, the present invention provides a preprocessing means that performs preprocessing, including noise reduction, on an audio signal acquired using an audio input means, and then uses a feature extraction means that extracts feature quantities from the preprocessed audio signal. Furthermore, it includes a determination means that determines whether the audio was synthesized based on these feature quantities, and a display means that displays a warning according to the determination result. In addition, by using a segment extraction means to extract a specific audio section, or by performing the determination using a neural network, it is possible to perform determination with both accuracy and real-time performance.

[0006] "Voice input means" refers to electronic device components or software functions that acquire voice data from an external source and enable internal processing.

[0007] "Preprocessing means" refers to a function that removes noise and normalizes the acquired audio signal to prepare it for analysis.

[0008] "Feature extraction means" refers to a function that extracts important patterns and numerical indicators from audio signals and converts them into data as features.

[0009] "Determination means" refers to a system, particularly an algorithm or model, that has the function of determining whether speech is synthesized based on extracted features.

[0010] "Display means" refers to components or interfaces used to visually present warnings or information to the user.

[0011] "Section extraction means" refers to a part that has the function of identifying and separating meaningful audio sections from an audio signal.

[0012] A "neural network" refers to an artificial intelligence algorithm that learns from large amounts of data and has the ability to detect specific patterns and features. [Brief explanation of the drawing]

[0013] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Figure 11] This is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] This is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] This is a sequence diagram showing the processing flow of the data processing system in Example 2, which incorporates an emotion engine. [Figure 14] This is a sequence diagram showing the processing flow of the data processing system in Application Example 2, which combines an emotion engine. [Modes for carrying out the invention]

[0014] An example of an embodiment of the system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0015] First, the terms used in the following description will be explained.

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

[0017] In the following embodiments, the numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0018] In the following embodiments, the numbered storage is one or more non-volatile storage devices that store various programs, various parameters, and the like. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes, and the like.

[0019] 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).

[0020] 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."

[0021] [First Embodiment]

[0022] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0023] 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.

[0024] 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).

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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.

[0029] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0030] As shown in Figure 2, in the data processing device 12, a specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0031] 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.

[0032] 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.

[0033] 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".

[0034] This invention relates to a system for analyzing speech in real time on a mobile device such as a smartphone and determining whether or not it is synthesized speech. This system acquires speech input from the user using a speech input means. The terminal applies preprocessing, such as noise reduction, to the acquired speech signal to improve the accuracy of the analysis. The terminal uses a feature extraction means to extract features from the preprocessed speech signal.

[0035] Once the features are extracted, the terminal uses a determination mechanism to determine whether or not they are synthesized speech. This determination is performed using a neural network-based algorithm, employing advanced pattern recognition technology to identify the characteristics of synthesized speech. If it is determined to be synthesized speech, the terminal uses a display mechanism to show a warning to the user.

[0036] For example, during a call, the system installed in the device analyzes the other party's voice, and if its characteristics match a synthesized voice pattern, the device immediately displays a warning message on the user's screen such as, "This voice may have been synthesized." This allows the user to be aware of the risk of fraud and take appropriate action. The server can collect more data on the cloud and continuously train the model to improve the accuracy of the detection.

[0037] The following describes the processing flow.

[0038] Step 1:

[0039] The device uses its built-in microphone to acquire audio and stores that audio data in a buffer in real time. At this stage, the digital data of the audio is prepared.

[0040] Step 2:

[0041] The device performs noise reduction processing on the audio data. Specifically, it uses digital filters to reduce background noise and ambient sounds, improving the clarity of the audio.

[0042] Step 3:

[0043] The device performs Voice Activity Detection (VAD) to extract only the actual speech segments. This removes silent or noise-only segments, improving the efficiency of the analysis.

[0044] Step 4:

[0045] The device extracts features from the pre-processed audio. Specifically, it calculates Mel-frequency cepstrum coefficients (MFCCs) and other audio features to quantitatively represent the characteristics of the audio.

[0046] Step 5:

[0047] The device inputs features into a neural network-based synthesized speech detection model. The model analyzes the extracted features and determines whether the speech is synthesized.

[0048] Step 6:

[0049] The device interprets the detection result and displays a warning message to the user if there is a high probability that the voice is synthesized. For example, it might display "This voice may have been synthesized" on the screen to alert the user.

[0050] Step 7:

[0051] The server, located in the cloud, continuously updates the model based on collected audio data and features to improve the accuracy of the predictions. This process improves the overall performance of the system.

[0052] (Example 1)

[0053] 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."

[0054] In recent years, advancements in speech synthesis technology have made it possible to sophisticatedly forge voices. This poses a potential threat to reliable communication. Therefore, there is a need for technology that can analyze voice data in real time and distinguish between synthesized and live voices.

[0055] 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.

[0056] In this invention, the server includes a voice acquisition means, a signal preprocessing means, a feature analysis means, a voice determination means, and a notification means. This enables users to engage in safe and reliable voice communication.

[0057] A "voice acquisition method" is a means of collecting voices emitted by the user in real time and inputting them into the system.

[0058] "Signal preprocessing means" refers to means of applying noise reduction and filtering to the acquired audio signal to prepare it for analysis.

[0059] A "feature analysis means" is a means that numerically extracts specific characteristics from a pre-processed audio signal and provides a foundation for analyzing audio data.

[0060] A "speech judgment method" is a means for distinguishing between synthesized speech and live speech based on analyzed characteristics and making an accurate judgment.

[0061] "Notification means" refers to means for notifying the user of the judgment result made by the voice judgment means and for displaying necessary information and warnings.

[0062] A "section selection method" is a means of selecting and extracting specific sections from an audio signal, enabling more detailed analysis and investigation.

[0063] A "multilayer perceptron" is a type of artificial intelligence technology used in speech analysis, and is a type of neural network that enables advanced pattern recognition of speech signals.

[0064] This invention relates to a system for analyzing voice in real time on a mobile information terminal and determining whether it is synthesized speech. The terminal uses voice acquisition means to collect voice emitted by the user. When voice is input via the built-in microphone, the terminal acquires the voice signal and automatically starts processing it.

[0065] The terminal performs noise reduction using signal preprocessing. Specifically, it utilizes Fourier transforms and filtering techniques to analyze the audio signal in the frequency domain, removes unwanted noise components, and generates clean audio data suitable for analysis. This preprocessing improves the accuracy of subsequent analysis.

[0066] From the processed audio signal, the terminal extracts features using a feature analysis tool. These features include spectral characteristics of the audio, such as Mel-frequency cepstrum coefficients (MFCCs), and are represented as numerical vectors.

[0067] The speech recognition method determines whether a speech is synthesized based on the extracted features. This uses a neural network model employing a multilayer perceptron, which, trained on past training data, highly recognizes speech patterns.

[0068] If the detection result indicates synthesized speech, the user will be immediately notified via the device's notification system. For example, if the other party's voice is detected as synthesized speech during a call, a warning message stating "This voice may be synthesized" will appear on the user's screen. This prompts the user to be aware of the risk and take appropriate action if necessary.

[0069] The server utilizes a cloud infrastructure to continuously collect data, enabling stepwise learning by the generative AI model. This process allows the model to adapt to new synthesized speech patterns, leading to long-term improvements in speech recognition accuracy.

[0070] Examples of prompt messages include instructions given by the user to the device, such as "Start analyzing synthesized speech." Based on this prompt message, the device immediately begins speech analysis.

[0071] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0072] Step 1:

[0073] The user inputs voice using the microphone on their mobile device. The device captures this voice through a voice acquisition device. The input is the user's raw voice, and the output is a digital audio signal.

[0074] Step 2:

[0075] The terminal processes the audio signal using a signal preprocessing mechanism. Here, it receives the audio signal as input and performs Fourier transform and filtering to remove noise. As a result, a clear audio signal suitable for analysis is output.

[0076] Step 3:

[0077] The terminal extracts features from a pre-processed audio signal using a feature analysis method. It takes a pre-processed audio signal as input and calculates Mel-frequency cepstrum coefficients (MFCCs). As a result, a numerical vector based on the spectral characteristics of the audio is output.

[0078] Step 4:

[0079] The terminal uses a speech recognition tool to determine whether the speech is synthesized based on the extracted features. It receives a feature vector as input and applies a neural network model using a multilayer perceptron to perform the determination. The output is the determination result indicating whether the speech is synthesized or not.

[0080] Step 5:

[0081] The terminal notifies the user of the result based on its judgment using a notification mechanism. It receives the judgment result as input, and if it determines that the voice is synthesized, it displays a warning message on the screen stating, "This voice may have been synthesized." The output is the displayed warning message.

[0082] Step 6:

[0083] The server continuously collects data via a cloud infrastructure and updates the model using a generative AI model. It acquires audio data that has been identified as input and uses the model's learning algorithm to continuously improve it. The output is an improved neural network model.

[0084] (Application Example 1)

[0085] 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."

[0086] In today's society, where voice communication is increasing, social problems using fraudulent synthesized voices, such as those used in scams, are on the rise. Existing systems struggle to detect such fraudulent activities in real time, leaving users vulnerable to these scams. New technologies are needed to address this challenge.

[0087] 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.

[0088] In this invention, the server includes an audio input device, a preprocessing device for removing noise from acquired audio data, and a feature extraction device for extracting features from the preprocessed audio data. This makes it possible to accurately detect fraudulent synthesized speech in real time during voice communication and to display a warning to the user.

[0089] A "voice input device" is a device used to acquire voice data.

[0090] A "preprocessing device" is a device used to remove unwanted noise from acquired audio data and improve the quality of the audio data.

[0091] A "feature extraction device" is a device used to extract features from pre-processed audio data.

[0092] A "determination device" is a device used to determine whether or not a voice has been artificially generated based on the extracted features.

[0093] A "display device" is a device that presents a warning to the user based on the judgment result.

[0094] An "information display device" is a device that provides users with the results of voice data analysis in a visual format.

[0095] A "partial extraction device" is a device used to selectively extract specific portions of audio data.

[0096] A "machine learning algorithm" is a computational method that learns patterns and rules based on data and uses that learning to make predictions and judgments about unknown data.

[0097] This invention is a system that uses an audio input device, a preprocessing device, a feature extraction device, a determination device, a display device, and an information presentation device to determine in real time whether audio data is artificially generated.

[0098] The server is equipped with an audio input device, which is used to acquire audio data. The audio data acquired by the audio input device is then preprocessed to remove noise and improve sound quality. This preprocessing is performed by a software library used for audio data cleansing (e.g., Librosa). The preprocessed audio data is then sent to a feature extraction device, where features that represent the characteristics of the audio data are extracted. Feature extraction is performed using a machine learning framework such as TENSORFLOW® or PyTorch.

[0099] Data obtained from the feature extraction device is sent to a determination device, where it is determined, based on a machine learning algorithm, whether the audio data is artificially generated. This determination result is notified to the user via a display device. The display device refers to the display of a smartphone or smart glasses, and its role is to visually present the determination result.

[0100] For example, if a user receives a voice message during a call that says, "Hello, this is a phishing attempt," the detection device will determine that it is highly likely to be an artificial voice and display a warning on the display device such as, "This voice may have been synthesized." This allows the user to be aware of the risks of phishing and fraud.

[0101] An example of a prompt in a generative AI model is, "This phone call has been detected as synthesized speech. Please create a message to warn the user of this display."

[0102] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0103] Step 1:

[0104] The server acquires voice data from the user using a voice input device. This input data is stored in memory as a raw audio signal. The voice data acquisition process takes place through the terminal's microphone.

[0105] Step 2:

[0106] The server removes noise from the acquired audio data using a preprocessor. The input is the audio signal acquired in step 1, and the output is a clean audio signal with the noise removed. The Librosa library is used for this process. Preprocessing is performed to improve the quality of the audio signal and enhance the accuracy of subsequent analysis.

[0107] Step 3:

[0108] The server extracts features from audio data that has been preprocessed by a feature extractor. The input is the clean audio signal processed in step 2, and the output is a feature vector that describes the characteristics of the audio. This process is performed using either TensorFlow or PyTorch. Feature extraction reveals important patterns and attributes contained in the data.

[0109] Step 4:

[0110] The server determines whether the speech is artificially generated based on features extracted using the judgment device. The input is the feature quantities obtained in step 3, and the output is the judgment result indicating the possibility of synthesized speech. This judgment is performed using a machine learning algorithm, enabling highly accurate speech identification.

[0111] Step 5:

[0112] The server notifies the user of the judgment result via a display device. The input is the judgment result obtained in step 4, and the output is the warning message presented to the user. This process allows the user to communicate with confidence.

[0113] Step 6:

[0114] The server uses an information display device to provide the user with a more detailed visual representation of the audio data's analysis results. The input is the warning message from step 5, and the output is information converted into a user-friendly format. For example, this may include displaying the message "This audio may have been synthesized" on the screen.

[0115] 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.

[0116] This invention relates to a system for mobile devices that not only analyzes voice in real time to determine if it is synthesized speech, but also analyzes the user's emotions. This system aims to enhance user convenience and safety by combining multiple functions.

[0117] First, the terminal acquires the user's voice via a voice input device. The acquired voice data is then de-noised and clarified by a pre-processing device. This makes the voice signal easier to analyze.

[0118] Next, the feature extraction means extracts features from the pre-processed audio signal. These features are provided to the determination means for synthesized speech detection. The determination means uses a neural network to analyze the possibility of synthesized speech from the input audio features and make a determination.

[0119] In addition, using a newly integrated emotion engine, the device analyzes emotions from the user's voice. It extracts emotional characteristics contained in the voice and uses the results to improve user safety and convenience.

[0120] As a concrete example, consider a situation where the user is on a call and this system analyzes the other party's voice and determines it to be synthesized speech. If the emotion engine determines that the user is in a state of tension, the device will likely take measures to make the warning message more prominent.

[0121] Furthermore, the data collected on the server is continuously analyzed and used to improve the overall system accuracy and update individual terminals. Learning using this data further improves the accuracy of synthesized speech recognition and emotion recognition, providing users with a safer and more reliable service.

[0122] The following describes the processing flow.

[0123] Step 1:

[0124] The device acquires the user's voice using a voice input method. The voice data is stored in a buffer in real time and prepared for subsequent processing.

[0125] Step 2:

[0126] The audio data acquired by the terminal undergoes preprocessing, including noise reduction. Specifically, filtering is performed to reduce ambient noise and prepare the audio signal for analysis.

[0127] Step 3:

[0128] The terminal extracts features from the pre-processed audio signal using a feature extraction method. It analyzes the time-frequency characteristics of the audio signal and calculates features such as Mel-frequency cepstrum coefficients (MFCCs).

[0129] Step 4:

[0130] The device inputs extracted features into a neural network-based classification system to analyze and determine whether the speech is synthesized. The model has learned the characteristics of synthesized speech, enabling highly accurate classification.

[0131] Step 5:

[0132] The device uses an emotion engine to analyze the user's emotions from their voice. It extracts emotional characteristics based on the intonation, tempo, and volume changes of the voice, and evaluates the user's current emotional state.

[0133] Step 6:

[0134] The device integrates the results of the detection method and the emotion engine and displays a warning through the user interface. For example, if it is determined that the voice is synthesized, and the emotion engine recognizes that the user is feeling tense, the device will present a more emphasized warning message.

[0135] Step 7:

[0136] The server collects data sent from the terminals and analyzes it for continuous model updates and accuracy improvements. Through this process, the entire system evolves, and the user experience improves.

[0137] (Example 2)

[0138] 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".

[0139] In recent years, advances in speech synthesis technology have led to problems with fraud and the spread of misinformation using synthesized voices. Furthermore, because the emotions contained in the voice influence the judgment of trustworthiness and urgency, there is a need to analyze the user's emotional state and provide appropriate warnings. However, conventional technologies have made it difficult to solve these problems simultaneously.

[0140] 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.

[0141] In this invention, the server includes a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism for extracting features, a judgment mechanism for synthesized voice judgment and emotion analysis, and a presentation mechanism for displaying warnings. This enables real-time determination of the reliability of voice information, analysis of the user's emotional state, and highly secure communication.

[0142] A "voice acquisition device" is an input device that collects voice information from the user.

[0143] A "pre-processing mechanism for noise reduction" is a mechanism that removes unwanted background noise from audio information and performs processing to clarify the signal.

[0144] A "feature extraction mechanism" is a mechanism that has the function of extracting important data characteristics for speech analysis from pre-processed speech information.

[0145] A "judgment mechanism" is an element of a system that makes judgments based on extracted features to evaluate whether or not speech synthesis has occurred and the emotional state of the speech.

[0146] A "presentation mechanism" is a device that conveys warnings or information to the user visually or audibly, based on the results of a judgment mechanism.

[0147] An "emotion analysis mechanism" is a device that uses technology to analyze a speaker's emotional state based on voice information.

[0148] The "Safety Assurance Mechanism" is a system that takes measures to enhance user safety based on the results of synthesized speech and emotion analysis.

[0149] This invention is a system that improves the security and convenience of communications by evaluating the reliability of voice information in real time and analyzing the user's emotions. This system mainly consists of a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism, a judgment mechanism, an emotion analysis mechanism, a presentation mechanism, and a security assurance mechanism.

[0150] The terminal receives voice input from the user using a voice acquisition device. At this stage, microphones and other voice input devices are used. The acquired voice information is then de-noised by a pre-processing mechanism within the terminal. This process utilizes general digital signal processing software, which is a signal processing library.

[0151] From the clear and clean audio signal, de-noise-removed, the device uses a feature extraction mechanism to extract the features necessary for speech analysis. Specifically, these might include Mel-frequency cepstrum coefficients (MFCCs) and speech spectrograms. Software such as Scikit-learn and TensorFlow are used for these techniques.

[0152] Next, the terminal uses a determination mechanism to evaluate whether the speech is synthesized using the extracted features. This evaluation process uses a pre-trained generative AI model. An example of a prompt might be, "Determine whether this speech is synthesized or not."

[0153] Simultaneously, the device uses an emotion analysis mechanism to analyze the user's emotional state contained in the audio. For example, a prompt such as "Identify the emotional state of the current audio" might be used. Based on the analysis results, if the user is in a specific emotional state, such as being tense, the device will issue a warning or notification through a notification mechanism. This function utilizes an interface that provides alerts and visual notifications.

[0154] Furthermore, the server continuously analyzes the collected voice information and analysis results to improve the overall accuracy of the system. Distributed data processing platforms such as Hadoop and Spark can be used for this data analysis. The server then uses this data to retrain the generative AI model, further improving the accuracy of voice recognition and sentiment analysis. As a result, the terminal can provide better service to users through the updated model.

[0155] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0156] Step 1:

[0157] The terminal receives voice input from the user using a voice acquisition device. This input is raw voice data and includes ambient noise. The output is a temporarily stored audio file. The terminal prepares this collected voice data for the next processing step.

[0158] Step 2:

[0159] The terminal uses a pre-processing mechanism to remove noise from the acquired audio data. The input is the audio file obtained in step 1. As part of the data processing, background noise is reduced using digital filtering technology. As a result, the output is a cleared audio signal, making subsequent analysis easier.

[0160] Step 3:

[0161] The terminal extracts features from the denoised audio signal via a feature extraction mechanism. The input for this step is the pre-processed audio signal. Specifically, Mel-frequency cepstrum coefficients (MFCCs) are calculated. The output is a feature vector, which is used in the next decision step.

[0162] Step 4:

[0163] The terminal uses a determination mechanism to determine if the speech is synthesized, using the feature vector extracted in step 3 as input. In this step, a generative AI model is applied, and calculations proceed based on the prompt "Determine whether this speech is synthesized or not." The output is a score indicating the likelihood that the speech is synthesized.

[0164] Step 5:

[0165] The device uses an emotion analysis mechanism to analyze the user's emotional state from the audio signal. The input for this process is again the feature vector obtained in step 3. The emotion analysis algorithm evaluates the intonation and pitch in the audio. The output is the user's emotion classification result.

[0166] Step 6:

[0167] The device issues warnings as needed, based on the results of its judgment and emotion analysis mechanisms. The inputs to this process are the judgment score and the emotion classification result. Specifically, a warning message is displayed on the screen via the notification mechanism, or an audible alert is emitted. The output is feedback to the user.

[0168] Step 7:

[0169] The server aggregates all judgment and sentiment analysis results sent from the terminals and stores them in a database. The input for this step is the analysis data sent from the terminals. Analysis is performed using a distributed data processing system, and the output is insights for improving the accuracy of the system. These insights are used to retrain the generative AI model and prepare for the next model update.

[0170] (Application Example 2)

[0171] 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".

[0172] In recent years, advancements in speech synthesis technology have made it difficult to distinguish between synthesized and natural speech. This poses a significant threat to privacy and security, particularly in information and communication. Furthermore, it is difficult to grasp the psychological state of the person you are speaking with in real time during a call, potentially leading to a decline in communication quality. To address these challenges, there is a need for a system that simultaneously performs voice authenticity testing and emotional state analysis.

[0173] 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.

[0174] In this invention, the server includes a voice acquisition means, a preprocessing means, a feature extraction means, an emotion analysis means, and a display means. This makes it possible to perform voice synthesis and emotion analysis simultaneously and present information in real time.

[0175] "Sound acquisition means" refers to a device or function for collecting external sound as input.

[0176] "Preprocessing means" refers to processing means performed to remove noise from collected audio information and clarify the information.

[0177] "Feature extraction means" refers to a device or mechanism for extracting specific feature information from pre-processed audio information.

[0178] "Determination means" refers to a device or method for determining whether or not speech has been synthesized based on extracted feature information.

[0179] "Emotional analysis means" refers to a device or mechanism for analyzing a speaker's emotional state from audio information.

[0180] "Display means" refers to a device or function for visually displaying information such as warning signals to the user based on the judgment result and emotional state.

[0181] "Section extraction means" refers to a device or function for extracting a specific section from audio information.

[0182] An "intelligent machine learning model" is an algorithm or method used for learning and prediction in the judgment of synthesized speech.

[0183] The system that realizes this application example operates using a terminal that employs various means. First, the terminal is equipped with an audio acquisition means that inputs ambient sound. The acquired audio information is then subjected to a preprocessing means, which performs noise reduction and clarifies the audio signal. This preprocessing improves the accuracy of subsequent analysis.

[0184] Next, the feature extraction means extracts feature information from the pre-processed audio data. This feature information includes, specifically, the pitch, rhythm, and spectral features of the audio. This information is provided to the determination means, where an intelligent machine learning model, such as a neural network utilizing TensorFlow or PyTorch, determines whether the audio is synthesized or not.

[0185] Furthermore, an emotion recognition engine is used to activate the emotion analysis mechanism and analyze the emotional state in the audio information. Through this process, emotions such as stress and anxiety are read from the user's way of speaking. The analysis results are sent to the display mechanism, and warning signals or cautionary messages are displayed on the user's screen as needed.

[0186] For example, when a user is on a call with a financial institution's customer service, if the system determines that the other party's voice is synthesized speech and also detects "stress" as the user's emotion, the terminal will display information on the screen saying, "Suspicious call detected, please be careful." This allows the user to immediately take precautions regarding the possibility of fraudulent communication.

[0187] As an example of a prompt to the generative AI model, we can use: "Retrieve feature information from the input speech and determine whether it is synthesized or natural speech. Also, analyze the speaker's emotional state and report any detected stress or anxiety." This prompt allows the AI ​​model to perform accurate analysis and provide highly practical results.

[0188] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0189] Step 1:

[0190] The device collects audio from the user's surroundings as input using an audio acquisition device. This audio data is then sent to the next processing step.

[0191] Step 2:

[0192] The terminal's preprocessing means performs noise reduction on the audio data obtained in step 1. It reduces background noise from the input audio signal and outputs clear audio information. This process improves the accuracy of the audio analysis.

[0193] Step 3:

[0194] The terminal uses a feature extraction means to extract feature information from the clear audio information obtained in step 2. This feature information includes the pitch and rhythm of the audio. The extracted data is provided to the determination means.

[0195] Step 4:

[0196] The device identification method uses an intelligent machine learning model to determine whether the speech is synthesized, based on the feature information extracted in step 3. This process utilizes a neural network to output the likelihood of the speech being synthesized.

[0197] Step 5:

[0198] The device's emotion analysis mechanism analyzes the speaker's emotional state from the audio information obtained in step 3. Using a generative AI model, it determines emotions such as stress and anxiety and outputs the results.

[0199] Step 6:

[0200] The terminal uses a display to show the user the judgment results and emotional state obtained in steps 4 and 5. If a warning is necessary, a warning message is displayed on the user's screen. This allows the user to immediately receive information about the safety of the call.

[0201] 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.

[0202] 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.

[0203] 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.

[0204] [Second Embodiment]

[0205] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0206] 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.

[0207] 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).

[0208] 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.

[0209] 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.

[0210] 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).

[0211] 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.

[0212] 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.

[0213] 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.

[0214] 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.

[0215] 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.

[0216] 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".

[0217] This invention relates to a system for analyzing speech in real time on a mobile device such as a smartphone and determining whether or not it is synthesized speech. This system acquires speech input from the user using a speech input means. The terminal applies preprocessing, such as noise reduction, to the acquired speech signal to improve the accuracy of the analysis. The terminal uses a feature extraction means to extract features from the preprocessed speech signal.

[0218] Once the features are extracted, the terminal uses a determination mechanism to determine whether or not they are synthesized speech. This determination is performed using a neural network-based algorithm, employing advanced pattern recognition technology to identify the characteristics of synthesized speech. If it is determined to be synthesized speech, the terminal uses a display mechanism to show a warning to the user.

[0219] For example, during a call, the system installed in the device analyzes the other party's voice, and if its characteristics match a synthesized voice pattern, the device immediately displays a warning message on the user's screen such as, "This voice may have been synthesized." This allows the user to be aware of the risk of fraud and take appropriate action. The server can collect more data on the cloud and continuously train the model to improve the accuracy of the detection.

[0220] The following describes the processing flow.

[0221] Step 1:

[0222] The device uses its built-in microphone to acquire audio and stores that audio data in a buffer in real time. At this stage, the digital data of the audio is prepared.

[0223] Step 2:

[0224] The device performs noise reduction processing on the audio data. Specifically, it uses digital filters to reduce background noise and ambient sounds, improving the clarity of the audio.

[0225] Step 3:

[0226] The device performs Voice Activity Detection (VAD) to extract only the actual speech segments. This removes silent or noise-only segments, improving the efficiency of the analysis.

[0227] Step 4:

[0228] The device extracts features from the pre-processed audio. Specifically, it calculates Mel-frequency cepstrum coefficients (MFCCs) and other audio features to quantitatively represent the characteristics of the audio.

[0229] Step 5:

[0230] The device inputs features into a neural network-based synthesized speech detection model. The model analyzes the extracted features and determines whether the speech is synthesized.

[0231] Step 6:

[0232] The device interprets the detection result and displays a warning message to the user if there is a high probability that the voice is synthesized. For example, it might display "This voice may have been synthesized" on the screen to alert the user.

[0233] Step 7:

[0234] The server, located in the cloud, continuously updates the model based on collected audio data and features to improve the accuracy of the predictions. This process improves the overall performance of the system.

[0235] (Example 1)

[0236] 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."

[0237] In recent years, advancements in speech synthesis technology have made it possible to sophisticatedly forge voices. This poses a potential threat to reliable communication. Therefore, there is a need for technology that can analyze voice data in real time and distinguish between synthesized and live voices.

[0238] 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.

[0239] In this invention, the server includes a voice acquisition means, a signal preprocessing means, a feature analysis means, a voice determination means, and a notification means. This enables users to engage in safe and reliable voice communication.

[0240] A "voice acquisition method" is a means of collecting voices emitted by the user in real time and inputting them into the system.

[0241] "Signal preprocessing means" refers to means of applying noise reduction and filtering to the acquired audio signal to prepare it for analysis.

[0242] A "feature analysis means" is a means that numerically extracts specific characteristics from a pre-processed audio signal and provides a foundation for analyzing audio data.

[0243] A "speech judgment method" is a means for distinguishing between synthesized speech and live speech based on analyzed characteristics and making an accurate judgment.

[0244] "Notification means" refers to means for notifying the user of the judgment result made by the voice judgment means and for displaying necessary information and warnings.

[0245] A "section selection method" is a means of selecting and extracting specific sections from an audio signal, enabling more detailed analysis and investigation.

[0246] A "multilayer perceptron" is a type of artificial intelligence technology used in speech analysis, and is a type of neural network that enables advanced pattern recognition of speech signals.

[0247] This invention relates to a system for analyzing voice in real time on a mobile information terminal and determining whether it is synthesized speech. The terminal uses voice acquisition means to collect voice emitted by the user. When voice is input via the built-in microphone, the terminal acquires the voice signal and automatically starts processing it.

[0248] The terminal performs noise reduction using signal preprocessing. Specifically, it utilizes Fourier transforms and filtering techniques to analyze the audio signal in the frequency domain, removes unwanted noise components, and generates clean audio data suitable for analysis. This preprocessing improves the accuracy of subsequent analysis.

[0249] From the processed audio signal, the terminal extracts features using a feature analysis tool. These features include spectral characteristics of the audio, such as Mel-frequency cepstrum coefficients (MFCCs), and are represented as numerical vectors.

[0250] The speech recognition method determines whether a speech is synthesized based on the extracted features. This uses a neural network model employing a multilayer perceptron, which, trained on past training data, highly recognizes speech patterns.

[0251] If the detection result indicates synthesized speech, the user will be immediately notified via the device's notification system. For example, if the other party's voice is detected as synthesized speech during a call, a warning message stating "This voice may be synthesized" will appear on the user's screen. This prompts the user to be aware of the risk and take appropriate action if necessary.

[0252] The server utilizes a cloud infrastructure to continuously collect data, enabling stepwise learning by the generative AI model. This process allows the model to adapt to new synthesized speech patterns, leading to long-term improvements in speech recognition accuracy.

[0253] Examples of prompt messages include instructions given by the user to the device, such as "Start analyzing synthesized speech." Based on this prompt message, the device immediately begins speech analysis.

[0254] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0255] Step 1:

[0256] The user inputs voice using the microphone on their mobile device. The device captures this voice through a voice acquisition device. The input is the user's raw voice, and the output is a digital audio signal.

[0257] Step 2:

[0258] The terminal processes the audio signal using a signal preprocessing mechanism. Here, it receives the audio signal as input and performs Fourier transform and filtering to remove noise. As a result, a clear audio signal suitable for analysis is output.

[0259] Step 3:

[0260] The terminal extracts features from a pre-processed audio signal using a feature analysis method. It takes a pre-processed audio signal as input and calculates Mel-frequency cepstrum coefficients (MFCCs). As a result, a numerical vector based on the spectral characteristics of the audio is output.

[0261] Step 4:

[0262] The terminal uses a speech recognition tool to determine whether the speech is synthesized based on the extracted features. It receives a feature vector as input and applies a neural network model using a multilayer perceptron to perform the determination. The output is the determination result indicating whether the speech is synthesized or not.

[0263] Step 5:

[0264] The terminal notifies the user of the result based on its judgment using a notification mechanism. It receives the judgment result as input, and if it determines that the voice is synthesized, it displays a warning message on the screen stating, "This voice may have been synthesized." The output is the displayed warning message.

[0265] Step 6:

[0266] The server continuously collects data via a cloud infrastructure and updates the model using a generative AI model. It acquires audio data that has been identified as input and continuously improves the model using its learning algorithm. The output is an improved neural network model.

[0267] (Application Example 1)

[0268] 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."

[0269] In today's society, where voice communication is increasing, social problems using fraudulent synthesized voices, such as those used in scams, are on the rise. Existing systems struggle to detect such fraudulent activities in real time, leaving users vulnerable to these scams. New technologies are needed to address this challenge.

[0270] 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.

[0271] In this invention, the server includes an audio input device, a preprocessing device for removing noise from acquired audio data, and a feature extraction device for extracting features from the preprocessed audio data. This makes it possible to accurately detect fraudulent synthesized speech in real time during voice communication and to display a warning to the user.

[0272] A "voice input device" is a device used to acquire voice data.

[0273] A "preprocessing device" is a device used to remove unwanted noise from acquired audio data and improve the quality of the audio data.

[0274] A "feature extraction device" is a device used to extract features from pre-processed audio data.

[0275] A "determination device" is a device used to determine whether or not a voice has been artificially generated based on the extracted features.

[0276] A "display device" is a device that presents a warning to the user based on the judgment result.

[0277] An "information display device" is a device that provides users with the results of voice data analysis in a visual format.

[0278] A "partial extraction device" is a device used to selectively extract specific portions of audio data.

[0279] A "machine learning algorithm" is a computational method that learns patterns and rules based on data and uses that learning to make predictions and judgments about unknown data.

[0280] This invention is a system that uses a voice input device, a preprocessing device, a feature extraction device, a determination device, a display device, and an information presentation device to determine in real time whether voice data is artificially generated.

[0281] The server is equipped with an audio input device, which is used to acquire audio data. The audio data acquired by the audio input device is then preprocessed to remove noise and improve sound quality. This preprocessing is performed by a software library used for audio data cleansing (e.g., Librosa). The preprocessed audio data is then sent to a feature extractor, where features that represent the characteristics of the audio data are extracted. Feature extraction is performed using a machine learning framework such as TensorFlow or PyTorch.

[0282] The data obtained from the feature extraction device is sent to the determination device, and based on the algorithm of machine learning, it is determined whether this voice data is artificially generated. This determination result is notified to the user via the display device. The display device refers to the display of a smartphone or smart glasses and plays a role in visually presenting the determination result.

[0283] For example, when the user receives a voice such as "Hello, this is a phishing attempt" during a call, the determination device determines that this is likely artificial voice and displays a warning such as "This voice may be synthesized" on the display device. Thereby, the user can be aware of the risks of phishing and fraud.

[0284] As an example of the prompt sentence in the generative AI model, there is "This phone voice has been detected as a synthetic voice. Please create a message to warn the user about this display."

[0285] The flow of the specific process in Application Example 1 will be described using FIG. 12.

[0286] Step 1:

[0287] The server acquires voice data from the user using the voice input device. This input data is stored in the memory as a raw voice signal. The process of acquiring voice data is performed through the microphone of the terminal.

[0288] Step 2:

[0289] The server removes noise from the acquired voice data using the preprocessing device. The input is the voice signal acquired in Step 1, and the output is a clean voice signal with noise removed. The Librosa library is used for this processing. The preprocessing is performed to improve the quality of the voice signal and enhance the subsequent analysis accuracy.

[0290] Step 3:

[0291] The server extracts features from audio data that has been preprocessed by a feature extractor. The input is the clean audio signal processed in step 2, and the output is a feature vector that describes the characteristics of the audio. This process is performed using either TensorFlow or PyTorch. Feature extraction reveals important patterns and attributes contained in the data.

[0292] Step 4:

[0293] The server determines whether the speech is artificially generated based on features extracted using the judgment device. The input is the feature quantities obtained in step 3, and the output is the judgment result indicating the possibility of synthesized speech. This judgment is performed using a machine learning algorithm, enabling highly accurate speech identification.

[0294] Step 5:

[0295] The server notifies the user of the judgment result via a display device. The input is the judgment result obtained in step 4, and the output is the warning message presented to the user. This process allows the user to communicate with confidence.

[0296] Step 6:

[0297] The server uses an information display device to provide the user with a more detailed visual representation of the audio data's analysis results. The input is the warning message from step 5, and the output is information converted into a user-friendly format. For example, this may include displaying the message "This audio may have been synthesized" on the screen.

[0298] 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.

[0299] This invention relates to a system for mobile devices that not only analyzes voice in real time to determine if it is synthesized speech, but also analyzes the user's emotions. This system aims to enhance user convenience and safety by combining multiple functions.

[0300] First, the terminal acquires the user's voice via a voice input device. The acquired voice data is then de-noised and clarified by a pre-processing device. This makes the voice signal easier to analyze.

[0301] Next, the feature extraction means extracts features from the pre-processed audio signal. These features are provided to the determination means for synthesized speech detection. The determination means uses a neural network to analyze the possibility of synthesized speech from the input audio features and make a determination.

[0302] In addition, using a newly integrated emotion engine, the device analyzes the user's emotions from their voice. It extracts emotional characteristics contained in the voice and uses the results to improve user safety and convenience.

[0303] As a concrete example, consider a situation where the user is on a call and this system analyzes the other party's voice and determines it to be synthesized speech. If the emotion engine determines that the user is in a state of tension, the device will likely take measures to make the warning message more prominent.

[0304] Furthermore, the data collected on the server is continuously analyzed and used to improve the overall accuracy of the system and update individual terminals. Through learning using this data, the accuracy of synthesized speech recognition and emotion recognition will be further improved, providing users with a safer and more secure service.

[0305] The following describes the processing flow.

[0306] Step 1:

[0307] The terminal acquires the user's voice using the voice input means. The voice data is stored in a buffer in real time for subsequent processing.

[0308] Step 2:

[0309] The terminal performs preprocessing including noise removal on the acquired voice data. Specifically, filtering is performed to reduce environmental noise and prepare the voice signal in a state suitable for analysis.

[0310] Step 3:

[0311] The terminal extracts feature quantities from the preprocessed voice signal using the feature extraction means. The time-frequency characteristics of the voice signal are analyzed, and features such as Mel-frequency cepstral coefficients (MFCC) are calculated.

[0312] Step 4:

[0313] The terminal inputs the extracted feature quantities into the neural network-based determination means to analyze and determine whether the voice is synthesized. The model has learned the characteristics of synthesized voices and can discriminate with high accuracy.

[0314] Step 5:

[0315] The terminal analyzes the emotion from the user's voice using the emotion engine. Based on the intonation, tempo, and volume changes of the voice, the features of the emotion are extracted to evaluate the user's current emotional state.

[0316] Step 6:

[0317] The terminal integrates the results of the determination means and the emotion engine and displays a warning through the user interface. For example, when it is determined that the voice is synthesized and the emotion engine recognizes that the user's emotion is tense, the terminal presents a more emphasized warning message.

[0318] Step 7:

[0319] The server collects data sent from the terminals and analyzes it for continuous model updates and accuracy improvements. Through this process, the entire system evolves, and the user experience improves.

[0320] (Example 2)

[0321] 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".

[0322] In recent years, advances in speech synthesis technology have led to problems with fraud and the spread of misinformation using synthesized voices. Furthermore, since the emotions contained in the voice influence the judgment of trustworthiness and urgency, there is a need to analyze the user's emotional state and provide appropriate warnings. However, conventional technologies have made it difficult to solve these problems simultaneously.

[0323] 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.

[0324] In this invention, the server includes a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism for extracting features, a judgment mechanism for synthesized voice judgment and emotion analysis, and a presentation mechanism for displaying warnings. This enables real-time determination of the reliability of voice information, analysis of the user's emotional state, and highly secure communication.

[0325] A "voice acquisition device" is an input device that collects voice information from the user.

[0326] A "pre-processing mechanism for noise reduction" is a mechanism that removes unwanted background noise from audio information and performs processing to clarify the signal.

[0327] A "feature extraction mechanism" is a mechanism that has the function of extracting important data characteristics for speech analysis from pre-processed speech information.

[0328] A "judgment mechanism" is an element of a system that makes judgments based on extracted features to evaluate whether or not speech synthesis has occurred and the emotional state of the speech.

[0329] A "presentation mechanism" is a device that conveys warnings or information to the user visually or audibly, based on the results of a judgment mechanism.

[0330] An "emotion analysis mechanism" is a device that uses technology to analyze a speaker's emotional state based on voice information.

[0331] The "Safety Assurance Mechanism" is a system that takes measures to enhance user safety based on the results of synthesized speech and emotion analysis.

[0332] This invention is a system that improves the security and convenience of communications by evaluating the reliability of voice information in real time and analyzing the user's emotions. This system mainly consists of a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism, a judgment mechanism, an emotion analysis mechanism, a presentation mechanism, and a security assurance mechanism.

[0333] The terminal receives voice input from the user using a voice acquisition device. At this stage, microphones and other voice input devices are used. The acquired voice information is then de-noised by a pre-processing mechanism within the terminal. This process utilizes general digital signal processing software, which is a signal processing library.

[0334] From the clear and clean audio signal, de-noise-removed, the device uses a feature extraction mechanism to extract the features necessary for speech analysis. Specifically, these might include Mel-frequency cepstrum coefficients (MFCCs) and speech spectrograms. Software such as Scikit-learn and TensorFlow are used for these techniques.

[0335] Next, the terminal uses a determination mechanism to evaluate whether the speech is synthesized using the extracted features. This evaluation process uses a pre-trained generative AI model. An example of a prompt might be, "Determine whether this speech is synthesized or not."

[0336] Simultaneously, the device uses an emotion analysis mechanism to analyze the user's emotional state contained in the audio. For example, a prompt such as "Identify the emotional state of the current audio" might be used. Based on the analysis results, if the user is in a specific emotional state, such as being tense, the device will issue a warning or notification through a notification mechanism. This function utilizes an interface that provides alerts and visual notifications.

[0337] Furthermore, the server continuously analyzes the collected voice information and analysis results to improve the overall accuracy of the system. Distributed data processing platforms such as Hadoop and Spark can be used for this data analysis. The server then uses this data to retrain the generative AI model, further improving the accuracy of voice recognition and sentiment analysis. As a result, the terminal can provide better service to users through the updated model.

[0338] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0339] Step 1:

[0340] The terminal receives voice input from the user using a voice acquisition device. This input is raw voice data and includes ambient noise. The output is a temporarily stored audio file. The terminal prepares this collected voice data for the next processing step.

[0341] Step 2:

[0342] The terminal uses a pre-processing mechanism to remove noise from the acquired audio data. The input is the audio file obtained in step 1. As part of the data processing, background noise is reduced using digital filtering technology. As a result, the output is a cleared audio signal, making subsequent analysis easier.

[0343] Step 3:

[0344] The terminal extracts features from the denoised audio signal via a feature extraction mechanism. The input for this step is the pre-processed audio signal. Specifically, Mel-frequency cepstrum coefficients (MFCCs) are calculated. The output is a feature vector, which is used in the next decision step.

[0345] Step 4:

[0346] The terminal uses a determination mechanism to determine if the speech is synthesized, using the feature vector extracted in step 3 as input. In this step, a generative AI model is applied, and calculations proceed based on the prompt "Determine whether this speech is synthesized or not." The output is a score indicating the likelihood that the speech is synthesized.

[0347] Step 5:

[0348] The device uses an emotion analysis mechanism to analyze the user's emotional state from the audio signal. The input for this process is again the feature vector obtained in step 3. The emotion analysis algorithm evaluates the intonation and pitch in the audio. The output is the user's emotion classification result.

[0349] Step 6:

[0350] The device issues warnings as needed, based on the results of its judgment and emotion analysis mechanisms. The inputs to this process are the judgment score and the emotion classification result. Specifically, a warning message is displayed on the screen via the notification mechanism, or an audible alert is emitted. The output is feedback to the user.

[0351] Step 7:

[0352] The server aggregates all judgment and sentiment analysis results sent from the terminals and stores them in a database. The input for this step is the analysis data sent from the terminals. Analysis is performed using a distributed data processing system, and the output is insights for improving the accuracy of the system. These insights are used to retrain the generative AI model and prepare for the next model update.

[0353] (Application Example 2)

[0354] 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."

[0355] In recent years, advancements in speech synthesis technology have made it difficult to distinguish between synthesized and natural speech. This poses a significant threat to privacy and security, particularly in information and communication. Furthermore, it is difficult to grasp the psychological state of the person you are speaking with in real time during a call, potentially leading to a decline in communication quality. To address these challenges, there is a need for a system that simultaneously performs voice authenticity testing and emotional state analysis.

[0356] 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.

[0357] In this invention, the server includes a voice acquisition means, a preprocessing means, a feature extraction means, an emotion analysis means, and a display means. This makes it possible to perform voice synthesis and emotion analysis simultaneously and present information in real time.

[0358] "Sound acquisition means" refers to a device or function for collecting external sound as input.

[0359] "Preprocessing means" refers to processing means performed to remove noise from collected audio information and clarify the information.

[0360] "Feature extraction means" refers to a device or mechanism for extracting specific feature information from pre-processed audio information.

[0361] "Determination means" refers to a device or method for determining whether or not speech has been synthesized based on extracted feature information.

[0362] "Emotional analysis means" refers to a device or mechanism for analyzing a speaker's emotional state from audio information.

[0363] "Display means" refers to a device or function for visually displaying information such as warning signals to the user based on the judgment result and emotional state.

[0364] "Section extraction means" refers to a device or function for extracting a specific section from audio information.

[0365] An "intelligent machine learning model" is an algorithm or method used for learning and prediction in the judgment of synthesized speech.

[0366] The system that realizes this application example operates using a terminal that employs various means. First, the terminal is equipped with an audio acquisition means that inputs ambient sound. The acquired audio information is then subjected to a preprocessing means, which performs noise reduction and clarifies the audio signal. This preprocessing improves the accuracy of subsequent analysis.

[0367] Next, the feature extraction means extracts feature information from the pre-processed audio data. This feature information includes, specifically, the pitch, rhythm, and spectral features of the audio. This information is provided to the determination means, where an intelligent machine learning model, such as a neural network utilizing TensorFlow or PyTorch, determines whether the audio is synthesized or not.

[0368] Furthermore, an emotion recognition engine is used to activate the emotion analysis mechanism and analyze the emotional state in the audio information. Through this process, emotions such as stress and anxiety are read from the user's way of speaking. The analysis results are sent to the display mechanism, and warning signals or cautionary messages are displayed on the user's screen as needed.

[0369] For example, when a user is on a call with a financial institution's customer service, if the system determines that the other party's voice is synthesized speech and also detects "stress" as the user's emotion, the terminal will display information on the screen saying, "Suspicious call detected, please be careful." This allows the user to immediately take precautions regarding the possibility of fraudulent communication.

[0370] As an example of a prompt to the generative AI model, we can use: "Retrieve feature information from the input speech and determine whether it is synthesized or natural speech. Also, analyze the speaker's emotional state and report any detected stress or anxiety." This prompt allows the AI ​​model to perform accurate analysis and provide highly practical results.

[0371] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0372] Step 1:

[0373] The device collects audio from the user's surroundings as input using an audio acquisition device. This audio data is then sent to the next processing step.

[0374] Step 2:

[0375] The terminal's preprocessing means performs noise reduction on the audio data obtained in step 1. It reduces background noise from the input audio signal and outputs clear audio information. This process improves the accuracy of the audio analysis.

[0376] Step 3:

[0377] The terminal uses a feature extraction means to extract feature information from the clear audio information obtained in step 2. This feature information includes the pitch and rhythm of the audio. The extracted data is provided to the determination means.

[0378] Step 4:

[0379] The device identification method uses an intelligent machine learning model to determine whether the speech is synthesized, based on the feature information extracted in step 3. This process utilizes a neural network to output the likelihood of the speech being synthesized.

[0380] Step 5:

[0381] The device's emotion analysis mechanism analyzes the speaker's emotional state from the audio information obtained in step 3. Using a generative AI model, it determines emotions such as stress and anxiety and outputs the results.

[0382] Step 6:

[0383] The terminal uses a display to show the user the judgment results and emotional state obtained in steps 4 and 5. If a warning is necessary, a warning message is displayed on the user's screen. This allows the user to immediately receive information about the safety of the call.

[0384] 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.

[0385] 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.

[0386] 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.

[0387] [Third Embodiment]

[0388] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0389] 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.

[0390] 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).

[0391] 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.

[0392] 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.

[0393] 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).

[0394] 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.

[0395] 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.

[0396] 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.

[0397] 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.

[0398] 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.

[0399] 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".

[0400] This invention relates to a system for analyzing speech in real time on a mobile device such as a smartphone and determining whether or not it is synthesized speech. This system acquires speech input from the user using a speech input means. The terminal applies preprocessing, such as noise reduction, to the acquired speech signal to improve the accuracy of the analysis. The terminal uses a feature extraction means to extract features from the preprocessed speech signal.

[0401] Once the features are extracted, the terminal uses a determination mechanism to determine whether or not they are synthesized speech. This determination is performed using a neural network-based algorithm, employing advanced pattern recognition technology to identify the characteristics of synthesized speech. If it is determined to be synthesized speech, the terminal uses a display mechanism to show a warning to the user.

[0402] For example, during a call, the system installed in the device analyzes the other party's voice, and if its characteristics match a synthesized voice pattern, the device immediately displays a warning message on the user's screen such as, "This voice may have been synthesized." This allows the user to be aware of the risk of fraud and take appropriate action. The server can collect more data on the cloud and continuously train the model to improve the accuracy of the detection.

[0403] The following describes the processing flow.

[0404] Step 1:

[0405] The device uses its built-in microphone to acquire audio and stores that audio data in a buffer in real time. At this stage, the digital data of the audio is prepared.

[0406] Step 2:

[0407] The device performs noise reduction processing on the audio data. Specifically, it uses digital filters to reduce background noise and ambient sounds, improving the clarity of the audio.

[0408] Step 3:

[0409] The device performs Voice Activity Detection (VAD) to extract only the actual speech segments. This removes silent or noise-only segments, improving the efficiency of the analysis.

[0410] Step 4:

[0411] The device extracts features from the pre-processed audio. Specifically, it calculates Mel-frequency cepstrum coefficients (MFCCs) and other audio features to quantitatively represent the characteristics of the audio.

[0412] Step 5:

[0413] The device inputs features into a neural network-based synthesized speech detection model. The model analyzes the extracted features and determines whether the speech is synthesized.

[0414] Step 6:

[0415] The device interprets the detection result and displays a warning message to the user if there is a high probability that the voice is synthesized. For example, it might display "This voice may have been synthesized" on the screen to alert the user.

[0416] Step 7:

[0417] The server, located in the cloud, continuously updates the model based on collected audio data and features to improve the accuracy of the predictions. This process improves the overall performance of the system.

[0418] (Example 1)

[0419] 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."

[0420] In recent years, advancements in speech synthesis technology have made it possible to sophisticatedly forge voices. This poses a potential threat to reliable communication. Therefore, there is a need for technology that can analyze voice data in real time and distinguish between synthesized and live voices.

[0421] 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.

[0422] In this invention, the server includes a voice acquisition means, a signal preprocessing means, a feature analysis means, a voice determination means, and a notification means. This enables users to engage in safe and reliable voice communication.

[0423] A "voice acquisition method" is a means of collecting voices emitted by the user in real time and inputting them into the system.

[0424] "Signal preprocessing means" refers to means of applying noise reduction and filtering to the acquired audio signal to prepare it for analysis.

[0425] A "feature analysis means" is a means that numerically extracts specific characteristics from a pre-processed audio signal and provides a foundation for analyzing audio data.

[0426] A "speech judgment method" is a means for distinguishing between synthesized speech and live speech based on analyzed characteristics and making an accurate judgment.

[0427] "Notification means" refers to means for notifying the user of the judgment result made by the voice judgment means and for displaying necessary information and warnings.

[0428] A "section selection method" is a means of selecting and extracting specific sections from an audio signal, enabling more detailed analysis and investigation.

[0429] A "multilayer perceptron" is a type of artificial intelligence technology used in speech analysis, and is a type of neural network that enables advanced pattern recognition of speech signals.

[0430] This invention relates to a system for analyzing voice in real time on a mobile information terminal and determining whether it is synthesized speech. The terminal uses voice acquisition means to collect voice emitted by the user. When voice is input via the built-in microphone, the terminal acquires the voice signal and automatically starts processing it.

[0431] The terminal performs noise reduction using signal preprocessing. Specifically, it utilizes Fourier transforms and filtering techniques to analyze the audio signal in the frequency domain, removes unwanted noise components, and generates clean audio data suitable for analysis. This preprocessing improves the accuracy of subsequent analysis.

[0432] From the processed audio signal, the terminal extracts features using a feature analysis tool. These features include spectral characteristics of the audio, such as Mel-frequency cepstrum coefficients (MFCCs), and are represented as numerical vectors.

[0433] The speech recognition method determines whether a speech is synthesized based on the extracted features. This uses a neural network model employing a multilayer perceptron, which, trained on past training data, highly recognizes speech patterns.

[0434] If the detection result indicates synthesized speech, the user will be immediately notified via the device's notification system. For example, if the other party's voice is detected as synthesized speech during a call, a warning message stating "This voice may be synthesized" will appear on the user's screen. This prompts the user to be aware of the risk and take appropriate action if necessary.

[0435] The server utilizes a cloud infrastructure to continuously collect data, enabling stepwise learning by the generative AI model. This process allows the model to adapt to new synthesized speech patterns, leading to long-term improvements in speech recognition accuracy.

[0436] Examples of prompt messages include instructions given by the user to the device, such as "Start analyzing synthesized speech." Based on this prompt message, the device immediately begins speech analysis.

[0437] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0438] Step 1:

[0439] The user inputs voice using the microphone on their mobile device. The device captures this voice through a voice acquisition device. The input is the user's raw voice, and the output is a digital audio signal.

[0440] Step 2:

[0441] The terminal processes the audio signal using a signal preprocessing mechanism. Here, it receives the audio signal as input and performs Fourier transform and filtering to remove noise. As a result, a clear audio signal suitable for analysis is output.

[0442] Step 3:

[0443] The terminal extracts features from a pre-processed audio signal using a feature analysis method. It takes a pre-processed audio signal as input and calculates Mel-frequency cepstrum coefficients (MFCCs). As a result, a numerical vector based on the spectral characteristics of the audio is output.

[0444] Step 4:

[0445] The terminal uses a speech recognition tool to determine whether the speech is synthesized based on the extracted features. It receives a feature vector as input and applies a neural network model using a multilayer perceptron to perform the determination. The output is the determination result indicating whether the speech is synthesized or not.

[0446] Step 5:

[0447] The terminal notifies the user of the result based on its judgment using a notification mechanism. It receives the judgment result as input, and if it determines that the voice is synthesized, it displays a warning message on the screen stating, "This voice may have been synthesized." The output is the displayed warning message.

[0448] Step 6:

[0449] The server continuously collects data via a cloud infrastructure and updates the model using a generative AI model. It acquires audio data that has been identified as input and continuously improves the model using its learning algorithm. The output is an improved neural network model.

[0450] (Application Example 1)

[0451] 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."

[0452] In today's society, where voice communication is increasing, social problems using fraudulent synthesized voices, such as those used in scams, are on the rise. Existing systems struggle to detect such fraudulent activities in real time, leaving users vulnerable to these scams. New technologies are needed to address this challenge.

[0453] 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.

[0454] In this invention, the server includes an audio input device, a preprocessing device for removing noise from acquired audio data, and a feature extraction device for extracting features from the preprocessed audio data. This makes it possible to accurately detect fraudulent synthesized speech in real time during voice communication and to display a warning to the user.

[0455] A "voice input device" is a device used to acquire voice data.

[0456] A "preprocessing device" is a device used to remove unwanted noise from acquired audio data and improve the quality of the audio data.

[0457] A "feature extraction device" is a device used to extract features from pre-processed audio data.

[0458] A "determination device" is a device used to determine whether or not a voice has been artificially generated based on the extracted features.

[0459] A "display device" is a device that presents a warning to the user based on the judgment result.

[0460] An "information display device" is a device that provides users with the results of voice data analysis in a visual format.

[0461] A "partial extraction device" is a device used to selectively extract specific portions of audio data.

[0462] A "machine learning algorithm" is a computational method that learns patterns and rules based on data and uses that learning to make predictions and judgments about unknown data.

[0463] This invention is a system that uses a voice input device, a preprocessing device, a feature extraction device, a determination device, a display device, and an information presentation device to determine in real time whether voice data is artificially generated.

[0464] The server is equipped with an audio input device, which is used to acquire audio data. The audio data acquired by the audio input device is then preprocessed to remove noise and improve sound quality. This preprocessing is performed by a software library used for audio data cleansing (e.g., Librosa). The preprocessed audio data is then sent to a feature extractor, where features that represent the characteristics of the audio data are extracted. Feature extraction is performed using a machine learning framework such as TensorFlow or PyTorch.

[0465] Data obtained from the feature extraction device is sent to a determination device, where it is determined, based on a machine learning algorithm, whether the audio data is artificially generated. This determination result is notified to the user via a display device. The display device refers to the display of a smartphone or smart glasses, and its role is to visually present the determination result.

[0466] For example, if a user receives a voice message during a call that says, "Hello, this is a phishing attempt," the detection device will determine that it is highly likely to be an artificial voice and display a warning on the display device such as, "This voice may have been synthesized." This allows the user to be aware of the risks of phishing and fraud.

[0467] An example of a prompt in a generative AI model is, "This phone call has been detected as synthesized speech. Please create a message to warn the user of this display."

[0468] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0469] Step 1:

[0470] The server acquires voice data from the user using a voice input device. This input data is stored in memory as a raw audio signal. The voice data acquisition process takes place through the terminal's microphone.

[0471] Step 2:

[0472] The server removes noise from the acquired audio data using a preprocessor. The input is the audio signal acquired in step 1, and the output is a clean audio signal with the noise removed. The Librosa library is used for this process. Preprocessing is performed to improve the quality of the audio signal and enhance the accuracy of subsequent analysis.

[0473] Step 3:

[0474] The server extracts features from audio data that has been preprocessed by a feature extractor. The input is the clean audio signal processed in step 2, and the output is a feature vector that describes the characteristics of the audio. This process is performed using either TensorFlow or PyTorch. Feature extraction reveals important patterns and attributes contained in the data.

[0475] Step 4:

[0476] The server determines whether the speech is artificially generated based on features extracted using the judgment device. The input is the feature quantities obtained in step 3, and the output is the judgment result indicating the possibility of synthesized speech. This judgment is performed using a machine learning algorithm, enabling highly accurate speech identification.

[0477] Step 5:

[0478] The server notifies the user of the judgment result via a display device. The input is the judgment result obtained in step 4, and the output is the warning message presented to the user. This process allows the user to communicate with confidence.

[0479] Step 6:

[0480] The server uses an information display device to provide the user with a more detailed visual representation of the audio data's analysis results. The input is the warning message from step 5, and the output is information converted into a user-friendly format. For example, this may include displaying the message "This audio may have been synthesized" on the screen.

[0481] 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.

[0482] This invention relates to a system for mobile devices that not only analyzes voice in real time to determine if it is synthesized speech, but also analyzes the user's emotions. This system aims to enhance user convenience and safety by combining multiple functions.

[0483] First, the terminal acquires the user's voice via a voice input device. The acquired voice data is then de-noised and clarified by a pre-processing device. This makes the voice signal easier to analyze.

[0484] Next, the feature extraction means extracts features from the pre-processed audio signal. These features are provided to the determination means for synthesized speech detection. The determination means uses a neural network to analyze the possibility of synthesized speech from the input audio features and make a determination.

[0485] In addition, using a newly integrated emotion engine, the device analyzes the user's emotions from their voice. It extracts emotional characteristics contained in the voice and uses the results to improve user safety and convenience.

[0486] As a concrete example, consider a situation where the user is on a call and this system analyzes the other party's voice and determines it to be synthesized speech. If the emotion engine determines that the user is in a state of tension, the device will likely take measures to make the warning message more prominent.

[0487] Furthermore, the data collected on the server is continuously analyzed and used to improve the overall accuracy of the system and update individual terminals. Through learning using this data, the accuracy of synthesized speech recognition and emotion recognition will be further improved, providing users with a safer and more secure service.

[0488] The following describes the processing flow.

[0489] Step 1:

[0490] The device acquires the user's voice using a voice input method. The voice data is stored in a buffer in real time and prepared for subsequent processing.

[0491] Step 2:

[0492] The audio data acquired by the terminal undergoes preprocessing, including noise reduction. Specifically, filtering is performed to reduce ambient noise and prepare the audio signal for analysis.

[0493] Step 3:

[0494] The terminal extracts features from the pre-processed audio signal using a feature extraction method. It analyzes the time-frequency characteristics of the audio signal and calculates features such as Mel-frequency cepstrum coefficients (MFCCs).

[0495] Step 4:

[0496] The device inputs extracted features into a neural network-based classification system to analyze and determine whether the speech is synthesized. The model has learned the characteristics of synthesized speech, enabling highly accurate classification.

[0497] Step 5:

[0498] The device uses an emotion engine to analyze the user's emotions from their voice. It extracts emotional characteristics based on the intonation, tempo, and volume changes of the voice, and evaluates the user's current emotional state.

[0499] Step 6:

[0500] The device integrates the results of the detection method and the emotion engine and displays a warning through the user interface. For example, if it is determined that the voice is synthesized, and the emotion engine recognizes that the user is feeling tense, the device will present a more emphasized warning message.

[0501] Step 7:

[0502] The server collects data sent from the terminals and analyzes it for continuous model updates and accuracy improvements. Through this process, the entire system evolves, and the user experience improves.

[0503] (Example 2)

[0504] 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."

[0505] In recent years, advances in speech synthesis technology have led to problems with fraud and the spread of misinformation using synthesized voices. Furthermore, since the emotions contained in the voice influence the judgment of trustworthiness and urgency, there is a need to analyze the user's emotional state and provide appropriate warnings. However, conventional technologies have made it difficult to solve these problems simultaneously.

[0506] 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.

[0507] In this invention, the server includes a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism for extracting features, a judgment mechanism for synthesized voice judgment and emotion analysis, and a presentation mechanism for displaying warnings. This enables real-time determination of the reliability of voice information, analysis of the user's emotional state, and highly secure communication.

[0508] A "voice acquisition device" is an input device that collects voice information from the user.

[0509] A "pre-processing mechanism for noise reduction" is a mechanism that removes unwanted background noise from audio information and performs processing to clarify the signal.

[0510] A "feature extraction mechanism" is a mechanism that has the function of extracting important data characteristics for speech analysis from pre-processed speech information.

[0511] A "judgment mechanism" is an element of a system that makes judgments based on extracted features to evaluate whether or not speech synthesis has occurred and the emotional state of the speech.

[0512] A "presentation mechanism" is a device that conveys warnings or information to the user visually or audibly, based on the results of a judgment mechanism.

[0513] An "emotion analysis mechanism" is a device that uses technology to analyze a speaker's emotional state based on voice information.

[0514] The "Safety Assurance Mechanism" is a system that takes measures to enhance user safety based on the results of synthesized speech and emotion analysis.

[0515] This invention is a system that improves the security and convenience of communications by evaluating the reliability of voice information in real time and analyzing the user's emotions. This system mainly consists of a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism, a judgment mechanism, an emotion analysis mechanism, a presentation mechanism, and a security assurance mechanism.

[0516] The terminal receives voice input from the user using a voice acquisition device. At this stage, microphones and other voice input devices are used. The acquired voice information is then de-noised by a pre-processing mechanism within the terminal. This process utilizes general digital signal processing software, which is a signal processing library.

[0517] From the clear and clean audio signal, de-noise-removed, the device uses a feature extraction mechanism to extract the features necessary for speech analysis. Specifically, these might include Mel-frequency cepstrum coefficients (MFCCs) and speech spectrograms. Software such as Scikit-learn and TensorFlow are used for these techniques.

[0518] Next, the terminal uses a determination mechanism to evaluate whether the speech is synthesized using the extracted features. This evaluation process uses a pre-trained generative AI model. An example of a prompt might be, "Determine whether this speech is synthesized or not."

[0519] Simultaneously, the device uses an emotion analysis mechanism to analyze the user's emotional state contained in the audio. For example, a prompt such as "Identify the emotional state of the current audio" might be used. Based on the analysis results, if the user is in a specific emotional state, such as being tense, the device will issue a warning or notification through a notification mechanism. This function utilizes an interface that provides alerts and visual notifications.

[0520] Furthermore, the server continuously analyzes the collected voice information and analysis results to improve the overall accuracy of the system. Distributed data processing platforms such as Hadoop and Spark can be used for this data analysis. The server then uses this data to retrain the generative AI model, further improving the accuracy of voice recognition and sentiment analysis. As a result, the terminal can provide better service to users through the updated model.

[0521] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0522] Step 1:

[0523] The terminal receives voice input from the user using a voice acquisition device. This input is raw voice data and includes ambient noise. The output is a temporarily stored audio file. The terminal prepares this collected voice data for the next processing step.

[0524] Step 2:

[0525] The terminal uses a pre-processing mechanism to remove noise from the acquired audio data. The input is the audio file obtained in step 1. As part of the data processing, background noise is reduced using digital filtering technology. As a result, the output is a cleared audio signal, making subsequent analysis easier.

[0526] Step 3:

[0527] The terminal extracts features from the denoised audio signal via a feature extraction mechanism. The input for this step is the pre-processed audio signal. Specifically, Mel-frequency cepstrum coefficients (MFCCs) are calculated. The output is a feature vector, which is used in the next decision step.

[0528] Step 4:

[0529] The terminal uses a determination mechanism to determine if the speech is synthesized, using the feature vector extracted in step 3 as input. In this step, a generative AI model is applied, and calculations proceed based on the prompt "Determine whether this speech is synthesized or not." The output is a score indicating the likelihood that the speech is synthesized.

[0530] Step 5:

[0531] The device uses an emotion analysis mechanism to analyze the user's emotional state from the audio signal. The input for this process is again the feature vector obtained in step 3. The emotion analysis algorithm evaluates the intonation and pitch in the audio. The output is the user's emotion classification result.

[0532] Step 6:

[0533] The device issues warnings as needed, based on the results of its judgment and emotion analysis mechanisms. The inputs to this process are the judgment score and the emotion classification result. Specifically, a warning message is displayed on the screen via the notification mechanism, or an audible alert is emitted. The output is feedback to the user.

[0534] Step 7:

[0535] The server aggregates all judgment and sentiment analysis results sent from the terminals and stores them in a database. The input for this step is the analysis data sent from the terminals. Analysis is performed using a distributed data processing system, and the output is insights for improving the accuracy of the system. These insights are used to retrain the generative AI model and prepare for the next model update.

[0536] (Application Example 2)

[0537] 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."

[0538] In recent years, advancements in speech synthesis technology have made it difficult to distinguish between synthesized and natural speech. This poses a significant threat to privacy and security, particularly in information and communication. Furthermore, it is difficult to grasp the psychological state of the person you are speaking with in real time during a call, potentially leading to a decline in communication quality. To address these challenges, there is a need for a system that simultaneously performs voice authenticity testing and emotional state analysis.

[0539] 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.

[0540] In this invention, the server includes a voice acquisition means, a preprocessing means, a feature extraction means, an emotion analysis means, and a display means. This makes it possible to perform voice synthesis and emotion analysis simultaneously and present information in real time.

[0541] "Sound acquisition means" refers to a device or function for collecting external sound as input.

[0542] "Preprocessing means" refers to processing means performed to remove noise from collected audio information and clarify the information.

[0543] "Feature extraction means" refers to a device or mechanism for extracting specific feature information from pre-processed audio information.

[0544] "Determination means" refers to a device or method for determining whether or not speech has been synthesized based on extracted feature information.

[0545] "Emotional analysis means" refers to a device or mechanism for analyzing a speaker's emotional state from audio information.

[0546] "Display means" refers to a device or function for visually displaying information such as warning signals to the user based on the judgment result and emotional state.

[0547] "Section extraction means" refers to a device or function for extracting a specific section from audio information.

[0548] An "intelligent machine learning model" is an algorithm or method used for learning and prediction in the judgment of synthesized speech.

[0549] The system that realizes this application example operates using a terminal that employs various means. First, the terminal is equipped with an audio acquisition means that inputs ambient sound. The acquired audio information is then subjected to a preprocessing means, which performs noise reduction and clarifies the audio signal. This preprocessing improves the accuracy of subsequent analysis.

[0550] Next, the feature extraction means extracts feature information from the pre-processed audio data. This feature information includes, specifically, the pitch, rhythm, and spectral features of the audio. This information is provided to the determination means, where an intelligent machine learning model, such as a neural network utilizing TensorFlow or PyTorch, determines whether the audio is synthesized or not.

[0551] Furthermore, an emotion recognition engine is used to activate the emotion analysis mechanism and analyze the emotional state in the audio information. Through this process, emotions such as stress and anxiety are read from the user's way of speaking. The analysis results are sent to the display mechanism, and warning signals or cautionary messages are displayed on the user's screen as needed.

[0552] For example, when a user is on a call with a financial institution's customer service, if the system determines that the other party's voice is synthesized speech and also detects "stress" as the user's emotion, the terminal will display information on the screen saying, "Suspicious call detected, please be careful." This allows the user to immediately take precautions regarding the possibility of fraudulent communication.

[0553] As an example of a prompt to the generative AI model, we can use: "Retrieve feature information from the input speech and determine whether it is synthesized or natural speech. Also, analyze the speaker's emotional state and report any detected stress or anxiety." This prompt allows the AI ​​model to perform accurate analysis and provide highly practical results.

[0554] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0555] Step 1:

[0556] The device collects audio from the user's surroundings as input using an audio acquisition device. This audio data is then sent to the next processing step.

[0557] Step 2:

[0558] The terminal's preprocessing means performs noise reduction on the audio data obtained in step 1. It reduces background noise from the input audio signal and outputs clear audio information. This process improves the accuracy of the audio analysis.

[0559] Step 3:

[0560] The terminal uses a feature extraction means to extract feature information from the clear audio information obtained in step 2. This feature information includes the pitch and rhythm of the audio. The extracted data is provided to the determination means.

[0561] Step 4:

[0562] The device identification method uses an intelligent machine learning model to determine whether the speech is synthesized, based on the feature information extracted in step 3. This process utilizes a neural network to output the likelihood of the speech being synthesized.

[0563] Step 5:

[0564] The device's emotion analysis mechanism analyzes the speaker's emotional state from the audio information obtained in step 3. Using a generative AI model, it determines emotions such as stress and anxiety and outputs the results.

[0565] Step 6:

[0566] The terminal uses a display to show the user the judgment results and emotional state obtained in steps 4 and 5. If a warning is necessary, a warning message is displayed on the user's screen. This allows the user to immediately receive information about the safety of the call.

[0567] 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.

[0568] 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.

[0569] 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.

[0570] [Fourth Embodiment]

[0571] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0572] 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.

[0573] 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).

[0574] 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.

[0575] 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.

[0576] 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).

[0577] 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.

[0578] 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.

[0579] 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.

[0580] 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.

[0581] 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.

[0582] 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.

[0583] 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".

[0584] This invention relates to a system for analyzing speech in real time on a mobile device such as a smartphone and determining whether or not it is synthesized speech. This system acquires speech input from the user using a speech input means. The terminal applies preprocessing, such as noise reduction, to the acquired speech signal to improve the accuracy of the analysis. The terminal uses a feature extraction means to extract features from the preprocessed speech signal.

[0585] Once the features are extracted, the terminal uses a determination mechanism to determine whether or not they are synthesized speech. This determination is performed using a neural network-based algorithm, employing advanced pattern recognition technology to identify the characteristics of synthesized speech. If it is determined to be synthesized speech, the terminal uses a display mechanism to show a warning to the user.

[0586] For example, during a call, the system installed in the device analyzes the other party's voice, and if its characteristics match a synthesized voice pattern, the device immediately displays a warning message on the user's screen such as, "This voice may have been synthesized." This allows the user to be aware of the risk of fraud and take appropriate action. The server can collect more data on the cloud and continuously train the model to improve the accuracy of the detection.

[0587] The following describes the processing flow.

[0588] Step 1:

[0589] The device uses its built-in microphone to acquire audio and stores that audio data in a buffer in real time. At this stage, the digital data of the audio is prepared.

[0590] Step 2:

[0591] The device performs noise reduction processing on the audio data. Specifically, it uses digital filters to reduce background noise and ambient sounds, improving the clarity of the audio.

[0592] Step 3:

[0593] The device performs Voice Activity Detection (VAD) to extract only the actual speech segments. This removes silent or noise-only segments, improving the efficiency of the analysis.

[0594] Step 4:

[0595] The device extracts features from the pre-processed audio. Specifically, it calculates Mel-frequency cepstrum coefficients (MFCCs) and other audio features to quantitatively represent the characteristics of the audio.

[0596] Step 5:

[0597] The device inputs features into a neural network-based synthesized speech detection model. The model analyzes the extracted features and determines whether the speech is synthesized.

[0598] Step 6:

[0599] The device interprets the detection result and displays a warning message to the user if there is a high probability that the voice is synthesized. For example, it might display "This voice may have been synthesized" on the screen to alert the user.

[0600] Step 7:

[0601] The server, located in the cloud, continuously updates the model based on collected audio data and features to improve the accuracy of the predictions. This process improves the overall performance of the system.

[0602] (Example 1)

[0603] 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".

[0604] In recent years, advancements in speech synthesis technology have made it possible to sophisticatedly forge voices. This poses a potential threat to reliable communication. Therefore, there is a need for technology that can analyze voice data in real time and distinguish between synthesized and live voices.

[0605] 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.

[0606] In this invention, the server includes a voice acquisition means, a signal preprocessing means, a feature analysis means, a voice determination means, and a notification means. This enables users to engage in safe and reliable voice communication.

[0607] A "voice acquisition method" is a means of collecting voices emitted by the user in real time and inputting them into the system.

[0608] "Signal preprocessing means" refers to means of applying noise reduction and filtering to the acquired audio signal to prepare it for analysis.

[0609] A "feature analysis means" is a means that numerically extracts specific characteristics from a pre-processed audio signal and provides a foundation for analyzing audio data.

[0610] A "speech judgment method" is a means for distinguishing between synthesized speech and live speech based on analyzed characteristics and making an accurate judgment.

[0611] "Notification means" refers to means for notifying the user of the judgment result made by the voice judgment means and for displaying necessary information and warnings.

[0612] A "section selection method" is a means of selecting and extracting specific sections from an audio signal, enabling more detailed analysis and investigation.

[0613] A "multilayer perceptron" is a type of artificial intelligence technology used in speech analysis, and is a type of neural network that enables advanced pattern recognition of speech signals.

[0614] This invention relates to a system for analyzing voice in real time on a mobile information terminal and determining whether it is synthesized speech. The terminal uses voice acquisition means to collect voice emitted by the user. When voice is input via the built-in microphone, the terminal acquires the voice signal and automatically starts processing it.

[0615] The terminal performs noise reduction using signal preprocessing. Specifically, it utilizes Fourier transforms and filtering techniques to analyze the audio signal in the frequency domain, removes unwanted noise components, and generates clean audio data suitable for analysis. This preprocessing improves the accuracy of subsequent analysis.

[0616] From the processed audio signal, the terminal extracts features using a feature analysis tool. These features include spectral characteristics of the audio, such as Mel-frequency cepstrum coefficients (MFCCs), and are represented as numerical vectors.

[0617] The speech recognition method determines whether a speech is synthesized based on the extracted features. This uses a neural network model employing a multilayer perceptron, which, trained on past training data, highly recognizes speech patterns.

[0618] If the detection result indicates synthesized speech, the user will be immediately notified via the device's notification system. For example, if the other party's voice is detected as synthesized speech during a call, a warning message stating "This voice may be synthesized" will appear on the user's screen. This prompts the user to be aware of the risk and take appropriate action if necessary.

[0619] The server utilizes a cloud infrastructure to continuously collect data, enabling stepwise learning by the generative AI model. This process allows the model to adapt to new synthesized speech patterns, leading to long-term improvements in speech recognition accuracy.

[0620] Examples of prompt messages include instructions given by the user to the device, such as "Start analyzing synthesized speech." Based on this prompt message, the device immediately begins speech analysis.

[0621] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0622] Step 1:

[0623] The user inputs voice using the microphone on their mobile device. The device captures this voice through a voice acquisition device. The input is the user's raw voice, and the output is a digital audio signal.

[0624] Step 2:

[0625] The terminal processes the audio signal using a signal preprocessing mechanism. Here, it receives the audio signal as input and performs Fourier transform and filtering to remove noise. As a result, a clear audio signal suitable for analysis is output.

[0626] Step 3:

[0627] The terminal extracts features from a pre-processed audio signal using a feature analysis method. It takes a pre-processed audio signal as input and calculates Mel-frequency cepstrum coefficients (MFCCs). As a result, a numerical vector based on the spectral characteristics of the audio is output.

[0628] Step 4:

[0629] The terminal uses a speech recognition tool to determine whether the speech is synthesized based on the extracted features. It receives a feature vector as input and applies a neural network model using a multilayer perceptron to perform the determination. The output is the determination result indicating whether the speech is synthesized or not.

[0630] Step 5:

[0631] The terminal notifies the user of the result based on its judgment using a notification mechanism. It receives the judgment result as input, and if it determines that the voice is synthesized, it displays a warning message on the screen stating, "This voice may have been synthesized." The output is the displayed warning message.

[0632] Step 6:

[0633] The server continuously collects data via a cloud infrastructure and updates the model using a generative AI model. It acquires audio data that has been identified as input and continuously improves the model using its learning algorithm. The output is an improved neural network model.

[0634] (Application Example 1)

[0635] 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".

[0636] In today's society, where voice communication is increasing, social problems using fraudulent synthesized voices, such as those used in scams, are on the rise. Existing systems struggle to detect such fraudulent activities in real time, leaving users vulnerable to these scams. New technologies are needed to address this challenge.

[0637] 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.

[0638] In this invention, the server includes an audio input device, a preprocessing device for removing noise from acquired audio data, and a feature extraction device for extracting features from the preprocessed audio data. This makes it possible to accurately detect fraudulent synthesized speech in real time during voice communication and to display a warning to the user.

[0639] A "voice input device" is a device used to acquire voice data.

[0640] A "preprocessing device" is a device used to remove unwanted noise from acquired audio data and improve the quality of the audio data.

[0641] A "feature extraction device" is a device used to extract features from pre-processed audio data.

[0642] A "determination device" is a device used to determine whether or not a voice has been artificially generated based on the extracted features.

[0643] A "display device" is a device that presents a warning to the user based on the judgment result.

[0644] An "information display device" is a device that provides users with the results of voice data analysis in a visual format.

[0645] A "partial extraction device" is a device used to selectively extract specific portions of audio data.

[0646] A "machine learning algorithm" is a computational method that learns patterns and rules based on data and uses that learning to make predictions and judgments about unknown data.

[0647] This invention is a system that uses a voice input device, a preprocessing device, a feature extraction device, a determination device, a display device, and an information presentation device to determine in real time whether voice data is artificially generated.

[0648] The server is equipped with an audio input device, which is used to acquire audio data. The audio data acquired by the audio input device is then preprocessed to remove noise and improve sound quality. This preprocessing is performed by a software library used for audio data cleansing (e.g., Librosa). The preprocessed audio data is then sent to a feature extractor, where features that represent the characteristics of the audio data are extracted. Feature extraction is performed using a machine learning framework such as TensorFlow or PyTorch.

[0649] Data obtained from the feature extraction device is sent to a determination device, where it is determined, based on a machine learning algorithm, whether the audio data is artificially generated. This determination result is notified to the user via a display device. The display device refers to the display of a smartphone or smart glasses, and its role is to visually present the determination result.

[0650] For example, if a user receives a voice message during a call that says, "Hello, this is a phishing attempt," the detection device will determine that it is highly likely to be an artificial voice and display a warning on the display device such as, "This voice may have been synthesized." This allows the user to be aware of the risks of phishing and fraud.

[0651] An example of a prompt in a generative AI model is, "This phone call has been detected as synthesized speech. Please create a message to warn the user of this display."

[0652] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0653] Step 1:

[0654] The server acquires voice data from the user using a voice input device. This input data is stored in memory as a raw audio signal. The voice data acquisition process takes place through the terminal's microphone.

[0655] Step 2:

[0656] The server removes noise from the acquired audio data using a preprocessor. The input is the audio signal acquired in step 1, and the output is a clean audio signal with the noise removed. The Librosa library is used for this process. Preprocessing is performed to improve the quality of the audio signal and enhance the accuracy of subsequent analysis.

[0657] Step 3:

[0658] The server extracts features from audio data that has been preprocessed by a feature extractor. The input is the clean audio signal processed in step 2, and the output is a feature vector that describes the characteristics of the audio. This process is performed using either TensorFlow or PyTorch. Feature extraction reveals important patterns and attributes contained in the data.

[0659] Step 4:

[0660] The server determines whether the speech is artificially generated based on features extracted using the judgment device. The input is the feature quantities obtained in step 3, and the output is the judgment result indicating the possibility of synthesized speech. This judgment is performed using a machine learning algorithm, enabling highly accurate speech identification.

[0661] Step 5:

[0662] The server notifies the user of the judgment result via a display device. The input is the judgment result obtained in step 4, and the output is the warning message presented to the user. This process allows the user to communicate with confidence.

[0663] Step 6:

[0664] The server uses an information display device to provide the user with a more detailed visual representation of the audio data's analysis results. The input is the warning message from step 5, and the output is information converted into a user-friendly format. For example, this may include displaying the message "This audio may have been synthesized" on the screen.

[0665] 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.

[0666] This invention relates to a system for mobile devices that not only analyzes voice in real time to determine if it is synthesized speech, but also analyzes the user's emotions. This system aims to enhance user convenience and safety by combining multiple functions.

[0667] First, the terminal acquires the user's voice via a voice input device. The acquired voice data is then de-noised and clarified by a pre-processing device. This makes the voice signal easier to analyze.

[0668] Next, the feature extraction means extracts features from the pre-processed audio signal. These features are provided to the determination means for synthesized speech detection. The determination means uses a neural network to analyze the possibility of synthesized speech from the input audio features and make a determination.

[0669] In addition, using a newly integrated emotion engine, the device analyzes the user's emotions from their voice. It extracts emotional characteristics contained in the voice and uses the results to improve user safety and convenience.

[0670] As a concrete example, consider a situation where the user is on a call and this system analyzes the other party's voice and determines it to be synthesized speech. If the emotion engine determines that the user is in a state of tension, the device will likely take measures to make the warning message more prominent.

[0671] Furthermore, the data collected on the server is continuously analyzed and used to improve the overall accuracy of the system and update individual terminals. Through learning using this data, the accuracy of synthesized speech recognition and emotion recognition will be further improved, providing users with a safer and more secure service.

[0672] The following describes the processing flow.

[0673] Step 1:

[0674] The device acquires the user's voice using a voice input method. The voice data is stored in a buffer in real time and prepared for subsequent processing.

[0675] Step 2:

[0676] The audio data acquired by the terminal undergoes preprocessing, including noise reduction. Specifically, filtering is performed to reduce ambient noise and prepare the audio signal for analysis.

[0677] Step 3:

[0678] The terminal extracts features from the pre-processed audio signal using a feature extraction method. It analyzes the time-frequency characteristics of the audio signal and calculates features such as Mel-frequency cepstrum coefficients (MFCCs).

[0679] Step 4:

[0680] The device inputs extracted features into a neural network-based classification system to analyze and determine whether the speech is synthesized. The model has learned the characteristics of synthesized speech, enabling highly accurate classification.

[0681] Step 5:

[0682] The device uses an emotion engine to analyze the user's emotions from their voice. It extracts emotional characteristics based on the intonation, tempo, and volume changes of the voice, and evaluates the user's current emotional state.

[0683] Step 6:

[0684] The device integrates the results of the detection method and the emotion engine and displays a warning through the user interface. For example, if it is determined that the voice is synthesized, and the emotion engine recognizes that the user is feeling tense, the device will present a more emphasized warning message.

[0685] Step 7:

[0686] The server collects data sent from the terminals and analyzes it for continuous model updates and accuracy improvements. Through this process, the entire system evolves, and the user experience improves.

[0687] (Example 2)

[0688] 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".

[0689] In recent years, advances in speech synthesis technology have led to problems with fraud and the spread of misinformation using synthesized voices. Furthermore, since the emotions contained in the voice influence the judgment of trustworthiness and urgency, there is a need to analyze the user's emotional state and provide appropriate warnings. However, conventional technologies have made it difficult to solve these problems simultaneously.

[0690] 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.

[0691] In this invention, the server includes a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism for extracting features, a judgment mechanism for synthesized voice judgment and emotion analysis, and a presentation mechanism for displaying warnings. This enables real-time determination of the reliability of voice information, analysis of the user's emotional state, and highly secure communication.

[0692] A "voice acquisition device" is an input device that collects voice information from the user.

[0693] A "pre-processing mechanism for noise reduction" is a mechanism that removes unwanted background noise from audio information and performs processing to clarify the signal.

[0694] A "feature extraction mechanism" is a mechanism that has the function of extracting important data characteristics for speech analysis from pre-processed speech information.

[0695] A "judgment mechanism" is an element of a system that makes judgments based on extracted features to evaluate whether or not speech synthesis has occurred and the emotional state of the speech.

[0696] A "presentation mechanism" is a device that conveys warnings or information to the user visually or audibly, based on the results of a judgment mechanism.

[0697] An "emotion analysis mechanism" is a device that uses technology to analyze a speaker's emotional state based on voice information.

[0698] The "Safety Assurance Mechanism" is a system that takes measures to enhance user safety based on the results of synthesized speech and emotion analysis.

[0699] This invention is a system that improves the security and convenience of communications by evaluating the reliability of voice information in real time and analyzing the user's emotions. This system mainly consists of a voice acquisition device, a pre-processing mechanism for noise reduction, a feature extraction mechanism, a judgment mechanism, an emotion analysis mechanism, a presentation mechanism, and a security assurance mechanism.

[0700] The terminal receives voice input from the user using a voice acquisition device. At this stage, microphones and other voice input devices are used. The acquired voice information is then de-noised by a pre-processing mechanism within the terminal. This process utilizes general digital signal processing software, which is a signal processing library.

[0701] From the clear and clean audio signal, de-noise-removed, the device uses a feature extraction mechanism to extract the features necessary for speech analysis. Specifically, these might include Mel-frequency cepstrum coefficients (MFCCs) and speech spectrograms. Software such as Scikit-learn and TensorFlow are used for these techniques.

[0702] Next, the terminal uses a determination mechanism to evaluate whether the speech is synthesized using the extracted features. This evaluation process uses a pre-trained generative AI model. An example of a prompt might be, "Determine whether this speech is synthesized or not."

[0703] Simultaneously, the device uses an emotion analysis mechanism to analyze the user's emotional state contained in the audio. For example, a prompt such as "Identify the emotional state of the current audio" might be used. Based on the analysis results, if the user is in a specific emotional state, such as being tense, the device will issue a warning or notification through a notification mechanism. This function utilizes an interface that provides alerts and visual notifications.

[0704] Furthermore, the server continuously analyzes the collected voice information and analysis results to improve the overall accuracy of the system. Distributed data processing platforms such as Hadoop and Spark can be used for this data analysis. The server then uses this data to retrain the generative AI model, further improving the accuracy of voice recognition and sentiment analysis. As a result, the terminal can provide better service to users through the updated model.

[0705] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0706] Step 1:

[0707] The terminal receives voice input from the user using a voice acquisition device. This input is raw voice data and includes ambient noise. The output is a temporarily stored audio file. The terminal prepares this collected voice data for the next processing step.

[0708] Step 2:

[0709] The terminal uses a pre-processing mechanism to remove noise from the acquired audio data. The input is the audio file obtained in step 1. As part of the data processing, background noise is reduced using digital filtering technology. As a result, the output is a cleared audio signal, making subsequent analysis easier.

[0710] Step 3:

[0711] The terminal extracts features from the denoised audio signal via a feature extraction mechanism. The input for this step is the pre-processed audio signal. Specifically, Mel-frequency cepstrum coefficients (MFCCs) are calculated. The output is a feature vector, which is used in the next decision step.

[0712] Step 4:

[0713] The terminal uses a determination mechanism to determine if the speech is synthesized, using the feature vector extracted in step 3 as input. In this step, a generative AI model is applied, and calculations proceed based on the prompt "Determine whether this speech is synthesized or not." The output is a score indicating the likelihood that the speech is synthesized.

[0714] Step 5:

[0715] The device uses an emotion analysis mechanism to analyze the user's emotional state from the audio signal. The input for this process is again the feature vector obtained in step 3. The emotion analysis algorithm evaluates the intonation and pitch in the audio. The output is the user's emotion classification result.

[0716] Step 6:

[0717] The device issues warnings as needed, based on the results of its judgment and emotion analysis mechanisms. The inputs to this process are the judgment score and the emotion classification result. Specifically, a warning message is displayed on the screen via the notification mechanism, or an audible alert is emitted. The output is feedback to the user.

[0718] Step 7:

[0719] The server aggregates all judgment and sentiment analysis results sent from the terminals and stores them in a database. The input for this step is the analysis data sent from the terminals. Analysis is performed using a distributed data processing system, and the output is insights for improving the accuracy of the system. These insights are used to retrain the generative AI model and prepare for the next model update.

[0720] (Application Example 2)

[0721] 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".

[0722] In recent years, advancements in speech synthesis technology have made it difficult to distinguish between synthesized and natural speech. This poses a significant threat to privacy and security, particularly in information and communication. Furthermore, it is difficult to grasp the psychological state of the person you are speaking with in real time during a call, potentially leading to a decline in communication quality. To address these challenges, there is a need for a system that simultaneously performs voice authenticity testing and emotional state analysis.

[0723] 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.

[0724] In this invention, the server includes a voice acquisition means, a preprocessing means, a feature extraction means, an emotion analysis means, and a display means. This makes it possible to perform voice synthesis and emotion analysis simultaneously and present information in real time.

[0725] "Sound acquisition means" refers to a device or function for collecting external sound as input.

[0726] "Preprocessing means" refers to processing means performed to remove noise from collected audio information and clarify the information.

[0727] "Feature extraction means" refers to a device or mechanism for extracting specific feature information from pre-processed audio information.

[0728] "Determination means" refers to a device or method for determining whether or not speech has been synthesized based on extracted feature information.

[0729] "Emotional analysis means" refers to a device or mechanism for analyzing a speaker's emotional state from audio information.

[0730] "Display means" refers to a device or function for visually displaying information such as warning signals to the user based on the judgment result and emotional state.

[0731] "Section extraction means" refers to a device or function for extracting a specific section from audio information.

[0732] An "intelligent machine learning model" is an algorithm or method used for learning and prediction in the judgment of synthesized speech.

[0733] The system that realizes this application example operates using a terminal that employs various means. First, the terminal is equipped with an audio acquisition means that inputs ambient sound. The acquired audio information is then subjected to a preprocessing means, which performs noise reduction and clarifies the audio signal. This preprocessing improves the accuracy of subsequent analysis.

[0734] Next, the feature extraction means extracts feature information from the pre-processed audio data. This feature information includes, specifically, the pitch, rhythm, and spectral features of the audio. This information is provided to the determination means, where an intelligent machine learning model, such as a neural network utilizing TensorFlow or PyTorch, determines whether the audio is synthesized or not.

[0735] Furthermore, an emotion recognition engine is used to activate the emotion analysis mechanism and analyze the emotional state in the audio information. Through this process, emotions such as stress and anxiety are read from the user's way of speaking. The analysis results are sent to the display mechanism, and warning signals or cautionary messages are displayed on the user's screen as needed.

[0736] For example, when a user is on a call with a financial institution's customer service, if the system determines that the other party's voice is synthesized speech and also detects "stress" as the user's emotion, the terminal will display information on the screen saying, "Suspicious call detected, please be careful." This allows the user to immediately take precautions regarding the possibility of fraudulent communication.

[0737] As an example of a prompt to the generative AI model, we can use: "Retrieve feature information from the input speech and determine whether it is synthesized or natural speech. Also, analyze the speaker's emotional state and report any detected stress or anxiety." This prompt allows the AI ​​model to perform accurate analysis and provide highly practical results.

[0738] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0739] Step 1:

[0740] The device collects audio from the user's surroundings as input using an audio acquisition device. This audio data is then sent to the next processing step.

[0741] Step 2:

[0742] The terminal's preprocessing means performs noise reduction on the audio data obtained in step 1. It reduces background noise from the input audio signal and outputs clear audio information. This process improves the accuracy of the audio analysis.

[0743] Step 3:

[0744] The terminal uses a feature extraction means to extract feature information from the clear audio information obtained in step 2. This feature information includes the pitch and rhythm of the audio. The extracted data is provided to the determination means.

[0745] Step 4:

[0746] The device identification method uses an intelligent machine learning model to determine whether the speech is synthesized, based on the feature information extracted in step 3. This process utilizes a neural network to output the likelihood of the speech being synthesized.

[0747] Step 5:

[0748] The device's emotion analysis mechanism analyzes the speaker's emotional state from the audio information obtained in step 3. Using a generative AI model, it determines emotions such as stress and anxiety and outputs the results.

[0749] Step 6:

[0750] The terminal uses a display to show the user the judgment results and emotional state obtained in steps 4 and 5. If a warning is necessary, a warning message is displayed on the user's screen. This allows the user to immediately receive information about the safety of the call.

[0751] 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.

[0752] 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.

[0753] 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.

[0754] 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.

[0755] 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.

[0756] 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.

[0757] 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.

[0758] 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.

[0759] 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."

[0760] 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.

[0761] 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.

[0762] 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.

[0763] 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.

[0764] 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.

[0765] 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.

[0766] 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.

[0767] 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.

[0768] 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.

[0769] 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.

[0770] 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.

[0771] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0772] The following is further disclosed regarding the embodiments described above.

[0773] (Claim 1)

[0774] Voice input method,

[0775] A pre-processing means for removing noise from the acquired audio signal,

[0776] A feature extraction means for extracting features from a pre-processed audio signal,

[0777] A determination means for determining whether the speech was synthesized based on the extracted features,

[0778] A display means that displays a warning based on the judgment result,

[0779] A system that includes this.

[0780] (Claim 2)

[0781] The system according to claim 1, further comprising a segment extraction means for extracting a specific section of an audio signal.

[0782] (Claim 3)

[0783] The system according to claim 1, further comprising a determination means that uses a neural network to determine the synthesized speech.

[0784] "Example 1"

[0785] (Claim 1)

[0786] A means of acquiring sound,

[0787] A signal preprocessing means for removing noise from an audio signal,

[0788] A feature analysis means for extracting characteristics from a pre-processed audio signal,

[0789] A voice determination means that determines whether the voice is synthesized based on the analyzed characteristics,

[0790] A notification means that displays a notification based on the judgment result,

[0791] A system that includes this.

[0792] (Claim 2)

[0793] The system according to claim 1, further comprising a segment selection means for extracting specific segments of an audio signal.

[0794] (Claim 3)

[0795] The system according to claim 1, further comprising a speech determination means that uses a multilayer perceptron to determine synthesized speech.

[0796] "Application Example 1"

[0797] (Claim 1)

[0798] Voice input device and

[0799] A preprocessing device for removing noise from acquired audio data,

[0800] A feature extraction device that extracts features from pre-processed audio data,

[0801] A determination device that determines whether the voice was artificially generated based on the extracted features,

[0802] A display device that presents a warning based on the judgment result,

[0803] An information display device that visually provides the user with the results of the voice data analysis,

[0804] A system that includes this.

[0805] (Claim 2)

[0806] The system according to claim 1, further comprising a partial extraction device for extracting a specific portion of audio data.

[0807] (Claim 3)

[0808] The system according to claim 1, comprising a judgment device that uses a machine learning algorithm to judge artificially generated speech.

[0809] "Example 2 of combining an emotion engine"

[0810] (Claim 1)

[0811] A voice acquisition device,

[0812] A pre-processing mechanism that removes noise from the acquired audio information,

[0813] A feature extraction mechanism that extracts features from pre-processed audio information,

[0814] A determination mechanism that determines whether the speech was synthesized based on the extracted features,

[0815] A warning mechanism that presents a warning based on the judgment result,

[0816] An emotion analysis mechanism that analyzes the user's emotions,

[0817] A safety mechanism that provides additional warnings or notifications based on analyzed sentiment information,

[0818] A system that includes this.

[0819] (Claim 2)

[0820] The system according to claim 1, comprising a function for analyzing emotions contained in audio information.

[0821] (Claim 3)

[0822] The system according to claim 1, further comprising a mechanism that uses a computational model for determining synthesized speech and analyzing emotions.

[0823] "Application example 2 when combining with an emotional engine"

[0824] (Claim 1)

[0825] A means of acquiring sound,

[0826] A preprocessing means for removing noise from acquired audio information,

[0827] A feature extraction means for extracting feature information from pre-processed audio information,

[0828] A determination means for determining whether speech has been synthesized based on extracted feature information,

[0829] An emotion analysis method that analyzes emotional states from audio information,

[0830] A display means that displays a warning signal based on the judgment result and emotional state,

[0831] A system that includes this.

[0832] (Claim 2)

[0833] The system according to claim 1, further comprising a segment extraction means for extracting a specific segment of audio information.

[0834] (Claim 3)

[0835] The system according to claim 1, further comprising a determination means that uses an intelligent machine learning model to judge synthesized speech. [Explanation of Symbols]

[0836] 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. Voice input method, A pre-processing means for removing noise from the acquired audio signal, A feature extraction means for extracting features from a pre-processed audio signal, A determination means for determining whether the speech was synthesized based on the extracted features, A display means that displays a warning based on the judgment result, A system that includes this.

2. The system according to claim 1, further comprising a segment extraction means for extracting a specific section of an audio signal.

3. The system according to claim 1, further comprising a determination means that uses a neural network to determine the synthesized speech.

Images