31 results about "Speech applications" patented technology

A neural network audio codec system and related methods are provided. In one example, a method is provided comprising: obtaining speech audio to be encoded; applying the speech audio to an audio encoder that is part of a neural network audio codec system that includes the audio encoder and an audio decoder. The audio encoder and the audio decoder have been trained in an end-to-end manner. The speech audio is encoded with the audio encoder to generate embedding vectors that represent a snapshot of speech audio attributes over successive timeframes of the raw speech audio, and from the embedding vectors, codeword indices are generated to entries in a codebook. The codeword indices are then transmitted or stored for later retrieval and processing by the audio decoder. Further aspects include converting the speech audio to text based on decoding the codeword indices, translating the speech audio from a first language into a second language by mapping codeword indices from a first language to a second language, converting the speech audio prosody by mapping the codeword indices, text generation and discrimination tasks.Yet another aspect includes converting text to speech audio with a different prosody.

A neural network audio codec system and related methods are provided. In one example, a method is provided comprising: obtaining speech audio to be encoded; applying the speech audio to an audio encoder that is part of a neural network audio codec system that includes the audio encoder and an audio decoder. The audio encoder and the audio decoder have been trained in an end-to-end manner. The speech audio is encoded with the audio encoder to generate embedding vectors that represent a snapshot of speech audio attributes over successive timeframes of the raw speech audio, and from the embedding vectors, codeword indices are generated to entries in a codebook. The codeword indices are then transmitted or stored for later retrieval and processing by the audio decoder.

A method (500) includes receiving a training utterance (106) including a sequence of audio frames (102) and a corresponding ground-truth transcription (320) and truncating a suffix portion from the sequence of audio frames to provide a prefix sequence of the audio frames. The method also includes processing, using a speech recognition model (200), the prefix sequence of audio frames to provide a partial speech recognition hypothesis (180a) and processing, using a prefix embedding multi-head attention layer (170), a sequence of prefix encodings (210) to generate a prefix prompt embedding (175). The method also includes processing, using a LM (160), the partial speech recognition hypothesis conditioned on the prefix prompt embedding to generate a speculated speech recognition (SSR) hypothesis (180b) and determining a training loss (325) based on the SSR hypothesis and the ground-truth transcription. The method also includes fine-tuning parameters of the prefix embedding multi-head attention layer.

The invention relates to the technical field of intelligent voice recognition and analysis, and discloses an intelligent voice recognition and analysis system for offline recording. The system comprises a recording data acquisition module, a feature hierarchical analysis module, a semantic matching module, an intention decision module, a central processing unit and the like. The recording data acquisition module divides an original audio data stream according to a preset voiceprint classification rule; the feature hierarchical analysis module extracts different semantic feature sets; the semantic matching module maps the feature set into an intention classification label; the intention decision module calls a target execution scheme; and the central processing unit checks the parameters to generate a final control instruction. In addition, the system is further provided with a voice resource scheduling module, an instruction conversion module, a voice archiving module and the like which are used for equipment scheduling, instruction conversion and data storage. The system can effectively process offline complex recording data, accurately recognize voice intentions, improve the intelligent level of voice interaction, and is suitable for various offline voice application scenes.

Systems and methods for protecting a voice application communicably coupled to a large language model (LLM) are disclosed herein. An example method is performed by one or more processors of a computing system. The example method may include: receiving an audio transmission over a communications network from a computing device associated with a user of the voice application; providing a request to the LLM based on the audio transmission; receiving a response to the request from the LLM; and performing one or more preemptive actions based on anticipating an anomaly in at least one of the request or the response.

A system and method providing automated voice-based healthcare plan delivery. The system interfaces with the patient through a voice-controlled personal assistant appliance via the internet. Some embodiments are implemented as voice applications or skills within the eco-system of an existing voice-controlled personal assistant device / appliance (hereinafter VCPAD). The system and method allow a clinician to create, digitally input, and modify a healthcare plan for a patient using a simple computer / web interface and without the need of any software development or computer programming skills, and allows a patient to interface with and query that healthcare plan using a largely conventional voice-controlled digital assistant appliance simply and efficiently to extract information about his / her individual healthcare plan in the patient's own language.

Embodiments of the present application provide a group communication method, comprising: listening, by a first terminal device, to one or more speech application channels; the one or more speech application channels being used by a plurality of second terminal devices to send speech application information; in a case where one or more speech application information is received in the one or more speech application channels, determining a target speech device from the second terminal devices that sent the one or more speech application information; sending, to the plurality of second terminal devices, speech confirmation information in a speech confirmation channel; the speech confirmation information indicating the target speech device and a first communication parameter; and receiving, based on the first communication parameter, group service data sent by the target speech device. Embodiments of the present application also provide a group communication apparatus, device, and chip.

The application discloses an applied far-field voice wake-up method and system, a terminal and a storage medium. The method comprises the following steps: when it is detected that a user issues an application wake-up instruction, obtaining recording data collected by a target device according to the application wake-up instruction, and performing enhanced noise reduction processing on the recording data to obtain target recording data; confirming the type of the target device to obtain device type information, and performing architecture decision evaluation according to the device type information to obtain a target wake-up architecture; performing data processing on the target recording data according to the target wake-up architecture to obtain an application wake-up signal, and waking up a corresponding voice application according to the application wake-up signal. The application can quickly evaluate the best wake-up support structure by introducing a decision module, and realizes the reuse of a wake-up algorithm module by customizing a unified general interface.

The application relates to the technical field of intelligent speech recognition and analysis, and discloses an intelligent speech recognition and analysis system for offline recording. The system comprises a recording data acquisition module, a feature hierarchical analysis module, a semantic matching module, an intention decision module, a central processing unit and the like. The recording data acquisition module divides original audio data streams according to preset voiceprint classification rules; the feature hierarchical analysis module extracts different semantic feature sets; the semantic matching module maps the feature sets into intention classification labels; the intention decision module calls a target execution scheme; and the central processing unit verifies parameters to generate a final control instruction. In addition, the system is also provided with a voice resource scheduling module, an instruction conversion module and a voice archiving module, which are used for device scheduling, instruction conversion and data storage. The system can effectively process offline complex recording data, accurately recognize voice intentions, improve the intelligent level of voice interaction, and is suitable for various offline voice application scenarios.

A computer-implemented system for facilitating a text to speech application, comprising a user communications device 104 configured to communicate with a Large Language Model (LLM) or Large Multimodal Model (LMM) API, the user communications device (104) being configured to: in real time, as soon as a text data stream has started to be received, use a speech emphasis handler module to segment the incoming text data into text data segments based on natural speech patterns and / or emphasis, and pass the text data segments, in order, as they are created, to a speech recognition module to transform each said text data segment into representative digital audio data; in real time, as said representative digital audio data is generated, start feeding said digital audio data to an audio buffer and store it therein; and deliver, in real time and in order, said audio segments to an audio playback device.

Methods and systems are provided for detecting and processing gestures, expressions (e.g., facial), tone and / or gestures of the user for the purpose of improving the quality and speed of interactions with computer-based systems. Such information may be detected by one or more sensors such as, for example, electromyography (EMG) sensors used to monitor and record electrical activity produced by muscles that are activated. Other sensor types may be used, such as optical, inertial measurement unit (IMU), or other types of bio-sensors. The system may use one or more sensors to detect speech alone or in combination with gestures, expressions (e.g., facial), tone and / or gestures of the user to provide input or control of the system.

To provide a system for supporting the intellectual training and growth of a child while reducing the labor of a parent.SOLUTION: The voice recognition system includes a means for acquiring voice uttered by a user, a voice recognition means for converting the acquired voice into text data, and a communication means for transmitting the converted text data. A system comprising: language analysis means for extracting a user's intention and emotional state; response generation means for generating an appropriate response based on the extracted intention and emotional state; communication means for receiving the generated response as text data; speech synthesis means for converting the received response text into voice data; and voice changer means for applying character voices to the converted voice data.SELECTED DRAWING: Figure 1

The invention discloses an iterative updating method, device and system for a voice application, a vehicle and a server, and relates to the technical field of voice recognition, the method is applied to a rail transit vehicle, and the method comprises the following steps: after the current voice application is started, obtaining an original voice data stream collected by a vehicle-mounted pickup; performing voice recognition on the original voice data stream by using a voice recognition model of the current voice application to obtain a voice recognition result; generating and storing a model iteration data set according to the voice recognition results and the recognition voice data corresponding to each voice recognition result; sending the model iteration data set to a cloud server; according to the method, the data required for updating the subsequent voice recognition model in the rail transit vehicle can be automatically acquired, so that data support is provided for iterative updating of the voice recognition model, rapid iterative updating of voice application in a rail transit scene is realized, and the engineering landing speed of a rail transit voice intelligent product is increased.

The application discloses a voice vehicle control full-link test method and device, electronic equipment and a storage medium. The method comprises the following steps: sending simulated voice instructions to a voice application in sequence, obtaining a controller state after the execution of the instructions is completed, judging whether the execution of the voice instructions is successful according to the controller state, and recording the execution time of the voice instructions when the execution of the voice instructions fails; synchronously intercepting a real vehicle CAN bus log and saving the real vehicle CAN bus log as an ASC file; according to the execution time of each voice instruction that fails to execute, whether the voice instruction corresponds to a CAN signal received by a comfort CAN and an information CAN within a set range of the execution time is inquired from the ASC file; and according to whether the CAN signal is received by the comfort CAN and the information CAN, a verification result is output. The voice vehicle control full-link test method and device, electronic equipment and storage medium provided by the application can automatically verify all voice vehicle control functions and output a verification result.

A system for detection of at least one designated wake-up word for at least one speech-enabled application. The system comprises at least one microphone; and at least one computer hardware processor configured to perform: receiving an acoustic signal generated by the at least one microphone at least in part as a result of receiving an utterance spoken by a speaker; obtaining information indicative of the speaker's identity; interpreting the acoustic signal at least in part by determining, using the information indicative of the speaker's identity and automated speech recognition, whether the utterance spoken by the speaker includes the at least one designated wake-up word; and interacting with the speaker based, at least in part, on results of the interpreting.

The present invention discloses a method and system for verifying electric power dispatching voice instructions based on logic verification, which relates to the technical field of voice recognition and comprises the following steps: obtaining pre-emphasized voice instructions and dividing them into frames; obtaining first instruction data, calculating the impact assessment data of the framing according to the first instruction data and a preset information loss factor, and performing data analysis to obtain the optimal framing length; applying a window function to each frame of voice under the optimal framing length, and performing a fast Fourier transform to obtain the energy distribution and balance data of the voice signal at different frequencies, and adjusting the window function length according to the balance data; performing data analysis on the energy distribution of the voice signal at different frequencies through a Mel filter to obtain a Mel spectrum, and performing a logarithmic conversion to obtain a Mel logarithmic spectrum; performing a cepstrum analysis on the Mel logarithmic spectrum to extract voice features; the present application improves the recognition ability of voice instructions by adjusting the frame length and the length of the window function.

Embodiments of the present application disclose an electronic device and a voice wake-up method, receiving voice data input by a user; performing voice processing on the voice data to obtain text information converted from the voice data; intercepting the text information according to a preset length to obtain a candidate text; dynamically matching a custom wake-up word and the candidate text to obtain a minimum edit distance between the custom wake-up word and the candidate text; calculating a confidence level according to the minimum edit distance; if the confidence level is greater than a threshold, waking up a voice application, the voice application being used to respond to a voice instruction input by the user; and if the confidence level is not greater than the threshold, not waking up the voice application. Embodiments of the present application can improve the accuracy of voice recognition and wake-up, avoid incorrect responses and execution of voice control instructions caused by user mis-wake-up of the voice application, reduce mis-wake-up and missed wake-up, improve wake-up rate and wake-up accuracy, and improve voice control function and user voice use experience.

The invention discloses a far-field voice wake-up method and system for an application, a terminal and a storage medium, and the method comprises the steps: obtaining recording data collected by a target device according to an application wake-up instruction when the application wake-up instruction sent by a user is detected, and carrying out the enhanced noise reduction processing of the recording data, and obtaining target recording data; performing type confirmation on the target device to obtain device type information, and performing architecture decision evaluation according to the device type information to obtain a target wake-up architecture; and performing data processing on the target recording data according to the target wake-up architecture to obtain an application wake-up signal, and waking up a corresponding voice application according to the application wake-up signal. The optimal wake-up support structure can be quickly evaluated by introducing the decision module, and the reuse of the wake-up algorithm module is realized by customizing a unified universal interface.

The invention relates to the technical field of artificial intelligence, can be applied to the technical field of medical health and finance, and discloses an environment perception speech synthesis method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a reference voice, a target environment voice, a target text and an environment perception value; inputting the reference voice, the target environment voice, the target text and the environment perception value into a synthesis model, converting the text into a vector sequence through a text coding layer by the synthesis model, and applying a mask strategy to the reference voice, processing the environment perception value and embedding a time step to regulate and control the intensity of the target environment voice, voice generation with environment perception is carried out, and environment perception voice is formed; and outputting the environment perception voice. By implementing the method provided by the embodiment of the invention, the naturalness and reality sense of environmental perception speech synthesis can be remarkably improved, and particularly, clearer and more natural speech communication experience is provided in remote medical consultation and financial customer service systems.

The invention provides a video saving method, electronic equipment, a chip system and a storage medium, and relates to the technical field of terminal equipment, and the method can be realized through a voice application and a video editing application: controlling the electronic equipment to generate a video through a voice instruction, and saving the video; in specific implementation, a voice application recognizes voice information of a user, and recognizes the voice information as a video generation instruction, a saving instruction or a saving cancelling instruction; the voice application sends the recognized instruction to the editing application; the editing application can execute corresponding operations of video generation, video saving and video saving canceling based on various received instructions; according to the method, a video editing function, a video storage function and the like can be opened to any application with a voice function; therefore, the video generation and storage processes are simplified; and user experience is improved.

The invention discloses a piezoelectric micro electro mechanical system acoustic sensor, which comprises a substrate, a device layer, an oxide layer, a piezoelectric layer and an electrode layer, sequentially laminating the device layer, the oxide layer, the piezoelectric layer and the electrode layer on the substrate; the device layer comprises a cantilever beam array, and the oxide layer is arranged on at least one part of the cantilever beam array; the substrate comprises a cavity, and the cavity penetrates through the substrate. The cantilever beam array can be utilized to form the acoustic sensor with high sensitivity, the acoustic sensor is suitable for various voice application scenes to realize sound sensing in most hearing ranges of human beings, the controllability of the sensitivity of the acoustic sensor is realized, and the acoustic sensor can be widely applied to the technical field of micro electro mechanical systems.

Systems and methods for decoding indented speech from neuronal activity in accordance with embodiments of the invention are illustrated. One embodiment includes a neuronal speech system for decoding intended speech from neuronal signals includes a neuronal signal recorder implanted into a user's brain, including a multielectrode array, controller circuitry, and a communication circuitry capable of transmitting data to a neuronal signal decoder, the neuronal signal decoder located externally from the user's body, including a processor, an input / output interface, and a memory, where the memory contains a neuronal speech application that directs the processor to obtain neuronal signal data from the neuronal signal recorder, where the neuronal signal data describes neuronal activity proximal to the implanted neuronal signal recorder, decode speech features from the neuronal signal data using a neuronal decoding model, construct a speech string from the decoded speech features; and output the constructed speech string via an output device.

The invention provides an expressway fatigue driving alarm method and system, the system comprises an infotainment controller, a driver monitoring module, a map application module and a voice application module, and the method comprises the steps that the driver monitoring module monitors the behavior state of a driver in real time, and when the driver is recognized to be in fatigue driving, the map application module sends an alarm to the map application module; the fatigue driving state is reported to an information entertainment controller; the information entertainment controller calls a map application module to judge whether a map is in a navigation state and a highway section; if the map is in a navigation state and in a highway section, searching a service area through a map application module; the voice application module broadcasts inquiry information and obtains confirmation information replied by the user; the map application module resets the navigation destination as the next nearest service area according to the confirmation information; the fatigue state of the user is accurately monitored, and the danger caused by fatigue driving of the user on the highway is avoided.

The invention relates to an audio acquisition method, device and equipment and a medium, and relates to the technical field of voice interaction, and the method comprises the steps: obtaining a target effective frame and an identification frame of a target audio; the target effective frame is an effective frame adjacent to the identification frame, and the identification frame comprises a cache frame and a supplementary frame; extracting a connection frame of the identification frame according to the amplitude and the zero-crossing rate of the target effective frame; and integrating the connection frame and all the effective frames to obtain an output audio. The head and tail missing parts of the effective voice are complemented, the output audio is ensured to be complete and coherent, and the effect of improving the recognition accuracy of the back-end voice application end is achieved.

For any application that processes speech, improving the quality of the feature vectors may improve the quality of the speech application. The quality of feature vectors may be improved by modifying a neural network architecture for computing feature vectors to allocate computational resources where they are more effective for learning and computing the feature vectors. Contextual feature vectors may be computed from feature vectors by using a parameterized downsampling operation that decreases a vector sequence rate, processing the downsampled vectors with a neural network, and using a parameterized upsampling operation that increases a vector sequence rate. For example, parameterized downsampling may decrease a vector sequence rate by a factor of two, a neural may require fewer computational resources since it operates with a lower vector sequence rate, and parameterized upsampling may then increase the vector sequence rate by a factor of two.

A distributed voice applications system includes a voice application rendering agent and at least one voice applications agent that is configured to provide voice applications to an individual user. A management system may control and direct the voice applications rendering agent to create voice applications that are personalized for individual users based on user characteristics, information about the environment in which the voice applications will be performed, prior user interactions and other information. The voice applications agent and components of customized voice applications may be resident on a local user device which includes a voice browser and speech recognition capabilities. The local device, voice applications rendering agent and management system may be interconnected via a communications network.

The invention discloses an intelligent terminal voice service input processing method and device and a terminal, and the method comprises the steps: obtaining the voice service input information of an intelligent terminal, and extracting the key information of the voice service input; information classification and abstraction are carried out on the extracted key information input by the intelligent terminal voice service, and the key information classification comprises a start event, an end event, audio data and / or text data; and according to the information classification determined by the key information input by the voice service of the intelligent terminal, controlling the independent processing flow of the corresponding classification to carry out voice recognition and voice application processing. According to the invention, the versatility and compatibility of voice service input of the television terminal can be improved, convenience is provided for television manufacturers and external integration parties, the quality and efficiency of the whole voice service are obviously improved, and convenience is provided for users.

The embodiment of the invention relates to the technical field of vehicle voice control, and discloses a vehicle voice interaction method and device and a computer readable storage medium, and the method comprises the steps: carrying out the voice recognition of an obtained first voice signal, and obtaining a first voice recognition result; processing the first voice recognition result and preset prompts corresponding to the plurality of interaction domains based on a large model to obtain intention structure data; obtaining vehicle end modal data corresponding to the intention structure data, and determining the voice instruction and voice resource information of a target language according to the vehicle end modal data and the intention structure data; and executing the voice instruction, loading a target voice processing resource corresponding to the target language when the voice resource information is not matched with the current voice processing resource, and outputting an execution confirmation voice of the target language based on the target voice processing resource. By applying the technical scheme of the invention, the accuracy and efficiency of voice control of the vehicle can be improved in a multilingual interaction environment.

A multi-voltage domain heterogeneous deep neural network (DNN) accelerator architecture includes an architecture that a) executes multiple DNN models simultaneously with different power-performance operating points; and b) improves the energy efficiency of near-memory computing applications by recycling leakage current of idle memories. The multi-voltage heterogenous DNN architecture may be implemented on battery operated or battery less edge devices with on-device intelligence executing applications including computer vision, augmented / virtual reality, face recognition, image processing, and speech applications.