# US 11,308,974 B2 · Target voice detection method and apparatus > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,308,974 B2 - Document ID: US11308974B2 - Status: Granted - Kind: B2 - Application number: 16/757,892 - Filed: 2018-07-16 - Published: 2020-10-29 - Granted: 2022-04-19 - Earliest claimed: 2017-10-23 - Assignees: Iflytek Co., Ltd. - Applicants: Iflytek Co., Ltd. - Inventors: Feng Ma; Haikun Wang; Zhiguo Wang; Guoping Hu - Web page: https://patentgrep.com/patent/US11308974B2 ## CPC - G10L 17/08: Use of distortion metrics or a particular distance between probe pattern and reference templates - G10L 21/0264: characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques - G10L 25/03: characterised by the type of extracted parameters - G10L 25/78: Detection of presence or absence of voice signals (switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10) - G10L 2021/02166: {Microphone arrays; Beamforming} - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract A target voice detection method and a target voice detection apparatus are provided. The method includes: receiving sound signals collected by a microphone array; performing a beamforming process on the sound signals to obtain beams in different directions; extracting a detection feature of each frame based on the sound signals and the beams in different directions; inputting an extracted detection feature of a current frame into a pre-constructed target voice detection model to obtain a model output result; and obtaining a target voice detection result of the current frame based on the model output result. ## Claims ### Claim 1 A target voice detection method, comprising: receiving sound signals collected by a microphone array; performing a beamforming process on each frame of the sound signals to obtain beams in different directions for the each frame; extracting a detection feature of the each frame based on the sound signals and the beams in the different directions; inputting an extracted detection feature of a current frame into a pre-constructed target voice detection model to obtain a model output result; obtaining a target voice detection result of the current frame based on the model output result; and performing target voice detection based on an intensity difference to obtain an intensity difference-based detection result, wherein the obtaining a target voice detection result of the current frame based on the model output result comprises: fusing the intensity difference-based detection result and the model output result to obtain the target voice detection result of the current frame. ### Claim 2 (depends on 1) The method according to claim 1, wherein a process of constructing the target voice detection model comprises: determining a topological structure of the target voice detection model; generating training data based on a clean voice and a simulated noise, and generating labeling information for labeling a target voice in the training data; extracting a detection feature of the training data; and obtaining parameters of the target voice detection model by performing training based on the detection feature and the labeling information. ### Claim 3 (depends on 1) The method according to claim 1, wherein the target voice detection model is a classification model or a regression model, and an output of the target voice detection model is an ideal binary mask or an ideal ratio mask for each frequency point of the current frame. ### Claim 4 (depends on 1) The method according to claim 1, wherein the detection feature comprises space dimension information, frequency dimension information, and time dimension information. ### Claim 5 (depends on 1) The method according to claim 1, wherein the extracting a detection feature of each frame based on the sound signals and the beams in different directions comprises: concatenating the sound signals collected by the microphone array and beam signals at each frequency point of each frame to obtain a multi-dimensional space vector; calculating a modulus of each element in the multi-dimensional space vector, and then concatenating moduli for all frequency points of each frame to obtain a multi-dimensional frequency vector containing space information; and performing frame expansion on the multi-dimensional frequency vector containing space information to obtain a multi-dimensional time vector containing space information and frequency information. ### Claim 6 (depends on 1) The method according to claim 1, wherein the performing target voice detection based on an intensity difference to obtain an intensity difference-based detection result comprises: obtaining a voice reference signal and a noise reference signal based on the beams in different directions; calculating a power of the voice reference signal and a power of the noise reference signal; calculating a power ratio of the power of the voice reference signal to the power of the noise reference signal; and obtaining the intensity difference-based detection result based on the power ratio. ### Claim 7 (depends on 1) A non-transitory computer readable storage medium comprising computer program codes, wherein the computer program codes, when executed by a computer unit, cause the computer unit to perform the target voice detection method according to claim 1. ### Claim 8 (depends on 1) A target voice detection device, comprising: a processor, a memory, and a system bus, wherein the processor and the memory are connected via the system bus; the memory stores one or more programs, wherein the one or more programs comprise instructions, and the instructions, when executed by the processor, cause the processor to perform the method according to claim 1. ### Claim 9 (depends on 1) A computer program product stored in a non-transitory computer readable storage medium, when executed on a terminal device, causing the terminal device to perform the method according to claim 1. ### Claim 10 A target voice detection apparatus, comprising: a signal receiving module, a beamforming module, a detection feature extracting module, a first detecting module, and a detection result outputting module, wherein the signal receiving module is configured to receive sound signals collected by a microphone array and output the sound signals to the beamforming module, the beamforming module is configured to perform a beamforming process on each frame of the inputted sound signals to obtain beams in different directions for the each frame, an input of the detection feature extracting module is connected to an output of the signal receiving module and an output of the beamforming module, and the detection feature extracting module is configured to extract a detection feature of the each frame based on the sound signals and the beams in the different directions, and output the extracted detection feature to the first detecting module, the first detecting module is configured to input a detection feature of a current frame extracted by the detection feature extracting module into a pre-constructed target voice detection model to obtain a model output result, and send the model output result to the detection result outputting module, and the detection result outputting module is configured to obtain a target voice detection result of the current frame based on the model output result, wherein the apparatus further comprises a second detecting module, having an input connected to an output of the beamforming module, configured to perform target voice detection based on an intensity difference to obtain an intensity difference-based detection result and send the intensity difference-based detection result to the detection result outputting module, and wherein the detection result outputting module is further configured to fuse the intensity difference-based detection result and the model output result to obtain the target voice detection result of the current frame. ### Claim 11 (depends on 10) The apparatus according to claim 10, further comprising: a model constructing module configured to construct the target voice detection model, wherein the model constructing module comprises: a structure design unit, configured to determine a topological structure of the target voice detection model, a training data processing unit, configured to generate training data based on a clean voice and a simulated noise and generate labeling information for labeling a target voice in the training data, a feature extraction unit, configured to extract a detection feature of the training data, and a training unit, configured to obtain parameters of the target voice detection model by performing training based on the detection feature and the labeling information. ### Claim 12 (depends on 10) The apparatus according to claim 10, wherein the target voice detection model is a classification model or a regression model. ### Claim 13 (depends on 10) The apparatus according to claim 10, wherein the second detecting module comprises: a reference signal obtaining unit, configured to obtain a voice reference signal and a noise reference signal based on the beams in different directions; a calculation unit, configured to calculate a power of the voice reference signal and a power of the noise reference signal and calculate a power ratio of the power of the voice reference signal to the power of the noise reference signal; and a detection result unit, configured to obtain the intensity difference-based detection result based on the power ratio.