195 results about "Time frequency transform" patented technology

The invention discloses a live detection method for insulation defects of high-voltage power equipment. The method comprises the following steps: firstly, synchronously arranging ultrahigh frequency sensors and acoustic emission sensors on a plurality of monitoring points on the surface of a gas insulated switchgear shell to form an array, and synchronously acquiring signals for filtering pretreatment; then, pulse events are extracted from the two types of signals respectively, and associated pulses in a time window are matched into matched pulse pairs representing the same discharge source; then, carrying out time-frequency transformation on the ultrahigh frequency and acoustic emission pulse waveform in each matched pulse pair, carrying out joint noise reduction by calculating a coherence coefficient between time-frequency distribution matrixes, and extracting a joint feature vector containing an energy ratio, a time parameter ratio and a frequency difference from the denoised time-frequency matrix; according to the invention, multi-source signals are fused, and high-sensitivity detection, high-precision positioning and high-accuracy identification of insulation defects are realized.

The invention relates to a welding seam abnormity identification method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring ultrasonic scanning data of a target welding seam; wherein the ultrasonic scanning data comprises an ultrasonic scanning result and corresponding time data; according to the time data, carrying out time-frequency conversion on the ultrasonic scanning data to obtain ultrasonic time-frequency data; inputting the ultrasonic time-frequency data into a preset anomaly detection model, and outputting to obtain an anomaly detection result; wherein the anomaly detection model is used for extracting target features according to input time-frequency data, and outputting a recognition result of the anomaly of the target welding seam based on the target features. By adopting the method, the abnormal welding seam can be identified more accurately, effective mapping from abstract time-frequency data to specific defect characteristics is realized, and a reliable basis is provided for welding seam quality evaluation.

The invention relates to the technical field of power equipment state monitoring, and discloses a power equipment fault detection method and device, equipment and a medium, and the method comprises the steps: obtaining a voiceprint signal of power equipment, carrying out the time-frequency transformation to obtain an original logarithmic Mel spectrogram, inputting a multi-scale context sensing auto-encoder model, and outputting a reconstructed spectrogram. According to the model, multi-scale long-range dependence features of voiceprints in time and frequency dimensions are respectively extracted by using a double-flow expansion convolutional network, and complete spectrum reconstruction is carried out based on the extracted features; and calculating an abnormal score based on a reconstruction difference degree between the original logarithmic Mel spectrogram and the reconstructed spectrogram, and when the abnormal score exceeds a dynamic threshold value, judging that the equipment has a fault. Compared with the prior art, the problems that weak fault features are difficult to extract and reconstruction details are fuzzy under strong background noise are solved, and high-robustness non-contact fault detection is achieved.

The invention relates to the technical field of signal processing, and provides a hidden weak signal feature recognition method and system, and the method comprises the steps: carrying out the time-frequency transformation of a to-be-recognized signal, and generating a time-frequency diagram; inputting the time-frequency graph into the feature recognition model, and extracting multi-scale features from the time-frequency graph through a backbone network; sending a feature map with the highest semantic hierarchy in the multi-scale features into a convolution attention module, and sequentially executing channel attention weighting and space attention weighting in the convolution attention module to obtain an enhanced feature map; fusing the enhanced feature map and other scale features in a feature fusion layer to obtain a fused feature map; and based on the fused feature map, identifying a weak signal through a detection head. Compared with the prior art, the method has the advantage that the recognition accuracy of weak signals in communication signals is greatly improved.

The invention relates to a power distribution network disturbance source identification method based on multi-modal feature fusion, and belongs to the technical field of refined disturbance monitoring and diagnosis of a smart power grid. The method comprises the steps of performing time-frequency transformation on extracted disturbance type current traveling wave data, generating a traveling wave panoramic oscillogram and a two-dimensional time-frequency diagram, fusing the traveling wave panoramic oscillogram and the two-dimensional time-frequency diagram into a three-channel time-frequency image, and extracting an image feature vector; carrying out Prony modal parameter fitting on the disturbance type current traveling wave data, extracting a modal parameter feature vector and projecting the modal parameter feature vector to a high-dimensional feature space; performing Clark transformation on the disturbance type current traveling wave data, calculating a zero-mode energy proportion feature and projecting the zero-mode energy proportion feature to a high-dimensional feature space; fusing the image feature vector, the projected physical feature vector and the projected modulus energy feature vector; and inputting the fused feature vector into a classifier, and outputting a category identification result of the power distribution network disturbance source. The technical problems that in the prior art, feature representation is single, physical interpretability is weak, and stable electric power fingerprints are difficult to form are solved.

The invention discloses an edge unmanned aerial vehicle identification method and system based on a bi-pass fusion-time domain self-attention pulse neural network. The method comprises the steps of dual-band signal preprocessing and time-frequency transformation, normalization and size adjustment, and target identification by a global feature extraction network. The system comprises a radio frequency signal preprocessing module and a global feature extraction network. According to the invention, high-efficiency and low-power-consumption identification of the unmanned aerial vehicle target is realized.

The invention relates to the technical field of directional pickup, and discloses a directional pickup method and device, computer equipment and a medium, and the method comprises the steps: obtaining time domain voice signals collected by at least two microphones, and carrying out the time-frequency conversion; calculating an observation phase difference of each frequency point, and determining a target theoretical phase difference according to a target pickup direction specified by a user and the geometric parameters of the microphone; encoding the target theoretical phase difference into a query vector, encoding the observation phase difference into a key vector, and generating a space confidence mask through a geometric cross attention mechanism; and filtering the frequency domain signal by using the mask, and outputting enhanced voice in a target direction through inverse time-frequency transformation. According to the invention, through an attention mechanism guided by physical prior, the problem of phase winding is effectively solved, and high-robustness and high-fidelity speech pickup in any direction is realized in a strong reverberation and low signal-to-noise ratio environment.

The invention discloses an unmanned aerial vehicle signal detection and identification method based on spectrum feature enhancement. The method comprises the following steps: firstly, carrying out time-frequency transformation on a received signal to obtain an original time-frequency graph; performing learning enhancement on the degenerated time-frequency graph by adopting a coding-decoding type deep neural network, and realizing noise suppression and structure recovery by combining pixel reconstruction, structural similarity and a texture perception loss function; and finally, performing target area detection and positioning on the enhanced time-frequency graph, directly outputting a structured result containing a time-frequency range, a category and confidence, and completing conversion from a frequency spectrum to a linkable engineering target. According to the unmanned aerial vehicle signal detection and recognition method based on spectrum feature enhancement, a spectrum feature enhancement mechanism is introduced before traditional spectrum analysis and feature recognition processing, and region-level detection and judgment are executed under the enhanced spectrum constraint condition; reliable discovery, positioning and identification of an unmanned aerial vehicle control link and an image transmission link in a complex electromagnetic environment are realized.

The invention relates to the technical field of impedance measurement, in particular to an impedance measurement method and system for a common-mode inductor, and the method comprises the steps: obtaining voltage and current signals of the common-mode inductor in a test circuit, and obtaining frequency domain response atlas data containing impedance characteristics through time-frequency transformation; carrying out non-linear segmentation division on the frequency axis to form a plurality of frequency segmentation regions; extracting the phase amplitude of each segmented region, carrying out clustering segmentation, and recognizing an effective impedance frequency band; analyzing phase synchronization characteristics in the effective impedance section, and determining a common-mode resonant frequency point through clustering analysis; and fusing multi-section frequency region division, effective impedance section and resonance point information to finish accurate measurement of the impedance characteristic of the common-mode inductor. According to the method, through non-linear scanning and phase information joint analysis, the recognition precision of impedance response key characteristics in a broadband range is improved, the capturing capacity of parasitic parameters and resonance behaviors is enhanced, and efficient and high-precision comprehensive evaluation of the impedance characteristics of the common-mode inductor is achieved.

The invention discloses a multi-source sensing fusion fault diagnosis method for intelligent equipment, and particularly relates to the technical field of current transformer monitoring. Collecting a residual current signal output by the mutual inductor according to a set sampling period, and generating corresponding time sequence data; carrying out time-frequency transformation processing on the signal to construct a two-dimensional atlas image; inputting the atlas into an atlas recognition model containing a convolutional neural network structure, extracting features and performing classification judgment; carrying out similarity or distribution deviation comparison on the current map characteristics and a preset normal category map, and judging whether a wiring abnormal state exists or not; if the real wiring fault is identified, alarm information is sent to a monitoring terminal; according to the method, intelligent, real-time and high-robustness identification of the wiring state of the mutual inductor can be realized, and the method is suitable for an electrical fire early warning system in a complex electrical environment.

The present application relates to the technical field of computer hardware fault diagnosis, in particular to a fault detection method for computer memory, comprising: applying a periodically changing voltage signal and temperature stress to the computer memory, and obtaining the dynamic current response sequence of the computer memory under dynamic combined stress; performing multi-scale time-frequency transformation on the current response sequence, and extracting a feature frequency component set of the memory under different voltage and temperature combination conditions; inputting the feature frequency component set into a pre-trained abnormal pattern recognition model, judging whether there is a potential fault risk according to the electrical characteristic change of the internal unit circuit of the memory; and generating a fault diagnosis report and providing a repair or replacement suggestion scheme according to the identified fault risk type and level. The present application can simulate the real working environment and realize early and accurate diagnosis of potential faults of the memory.

The invention provides a point-to-point communication method and system based on self-organizing direct connection, and the method comprises the steps: obtaining network state information in a plurality of time periods, obtaining a historical communication data set sequence, carrying out the construction of a communication topological graph, and carrying out the time sequence feature modeling, and obtaining a dynamic communication time-space diagram sequence; performing space-time correlation feature analysis and time sequence communication state evolution on the dynamic communication space-time diagram sequence to obtain an expected communication state data set corresponding to each communication link in a future period; determining a communication device group needing point-to-point communication, and performing multi-constraint adaptive path decision on a local topology link of the communication device group according to the expected communication data to obtain a standard communication link; and performing signal transmission among the communication equipment groups by using a standard communication link to obtain a received signal, performing time-frequency transformation, image denoising and feature enhancement on the received signal to obtain an enhanced frequency domain feature map, and performing inverse transformation on the enhanced frequency domain feature map to obtain an enhanced received signal.

This invention discloses a system and method for labeling automotive abnormal noise data, relating to the field of automotive NVH detection technology. The method includes: receiving raw signal data of automotive abnormal noise; performing time-frequency transformation on the raw signal to generate a time-frequency diagram; identifying key frequency ranges characterizing the abnormal noise features; then performing directional filtering to enhance the abnormal noise signal; identifying candidate time ranges for abnormal noise through an automatic detection algorithm; and outputting the final abnormal noise time range labels and key frequency ranges in a structured format after verification and correction. The system includes a data receiving module, a signal processing module, a display module, an intelligent labeling module, an audio playback module, a labeling output module, and an interaction module, achieving triple-assisted labeling through auditory, visual, and intelligent pre-selection methods. This solves the problems of low efficiency, low accuracy, and poor consistency in traditional manual listening methods, providing high-quality labeled data for deep learning classification of automotive abnormal noises.

The invention discloses an audio-visual joint detection and classification method for a faulty carrier roller of a belt conveyor, which belongs to the field of intelligent monitoring and fault diagnosis of the belt conveyor, and comprises the following steps: synchronously acquiring an image signal and an audio signal of the belt conveyor, and carrying out time sequence alignment processing to obtain synchronous image frames and audio clips; processing the audio clip by adopting a time-frequency transformation method to obtain a two-dimensional audio spectrogram; performing channel splicing on the image frames and the corresponding two-dimensional audio spectrograms, and inputting the image frames and the corresponding two-dimensional audio spectrograms into a single feature extraction network to obtain cross-modal fusion features; based on the cross-modal fusion features, pixel-level positioning and fault type identification of the fault carrier roller are carried out through a double-branch structure, and a segmentation mask and a fault type are obtained; and performing end-to-end joint training on the feature extraction network and the double-branch structure by adopting a multi-task joint loss function based on the label so as to synchronously optimize the positioning and classification tasks. According to the invention, the positioning precision of the fault sounding carrier roller is obviously improved.

The application provides a minimum information entropy criterion-based adaptive window length time-frequency transform algorithm, comprising: obtaining an input signal sequence, and calculating a correlation function of the signal; summing the correlation function to obtain S; setting multiple thresholds based on S; searching a peak width of the correlation function according to the thresholds to obtain the peak width corresponding to each threshold; taking the peak width corresponding to each threshold as a window length, calculating a time-frequency transform of the signal to obtain a plurality of time-frequency spectrums; calculating information entropy of the plurality of time-frequency spectrums respectively; selecting a minimum value in the information entropy; and taking a time-frequency spectrum corresponding to the minimum value as a time-frequency transform result. The window length can be adaptively adjusted, a time-frequency spectrum graph with better energy focusing performance is obtained, and more accurate extraction of signal components is realized.

The application discloses a kind of real-time detection method and system of food pollutant based on intelligent sensor, comprising: through the multimodal sensor array of deployment in food processing production line key station, sensor response signal is collected;Baseline response under the condition of no pollution is predicted by establishing dynamic baseline prediction model;Signal separation network based on variational decoupling autoencoder is constructed, residual signal is mapped to low-dimensional latent space by sharing encoder and is divided into baseline error subspace and pollutant signal subspace, respectively by two independent decoders reconstructing baseline prediction error signal and pollutant candidate signal;Multi-scale time-frequency transform is carried out to pollutant candidate signal to identify pollutant type and estimate concentration;Through distributed intelligent agent cooperation, pollution traceability and hierarchical early warning are carried out;Cross-scene adaptive calibration is realized using meta-learning framework.The application realizes the real-time decoupling of pollutant signal and matrix change signal in food processing process.

The application discloses a flywheel energy storage-based active quick-response frequency stabilization method and system for an isolated network, and relates to the field of power grid control.The method comprises the following steps: collecting a frequency signal of an isolated power grid in real time, fitting the frequency signal, and obtaining a smooth frequency value and a frequency change rate; when the absolute value of the frequency change rate exceeds a dynamic starting threshold, acquiring frequency deviation time series data of a preset time length to generate a disturbance event analysis window; extracting a frequency deviation signal in the disturbance event analysis window, performing time-frequency conversion on the frequency deviation signal, and generating a time-frequency spectrum diagram; based on the time-frequency spectrum diagram, decomposing the frequency deviation signal into a fundamental component and at least one overtone component, extracting a morphological feature parameter, and generating a disturbance image; setting a target state and a constraint condition of frequency recovery according to the disturbance image, determining a frequency recovery trajectory; and based on the frequency recovery trajectory, cyclically executing a control strategy until the power grid frequency is stabilized in a preset frequency range.

The application discloses a water supply pipeline leakage detection method based on limited samples, comprising: acquiring a water supply pipeline vibration signal and constructing a training data set; performing time-frequency transformation on the vibration signal, performing adaptive frequency band division based on random characteristic energy spectrum and maximum peak envelope segmentation of scale space representation, and applying continuous soft attenuation weight to obtain enhanced time-frequency features; generating a synthetic sample by using an improved denoising diffusion generative adversarial network to expand the training data set; constructing a multi-view contrast learning framework, including a time sequence feature extraction branch and a time-frequency feature extraction branch, performing self-supervised pre-training by using unlabeled samples, and jointly optimizing a time sequence contrast loss, a time-frequency contrast loss and a time sequence-time frequency consistency loss; fine-tuning the pre-training network by using labeled samples to obtain a leakage detection classification model; and inputting a signal to be detected into the model to output a leakage detection result. The application realizes high-precision and strong-robustness leakage detection under the conditions of strong noise, small samples and label scarcity.

The application provides a speech recognition method, system and device, which can be applied to the technical field of speech processing. The speech recognition method comprises the following steps: in response to a speech recognition instruction, obtaining original audio data, wherein the original audio data comprises a speech signal with background music noise; performing time-frequency conversion on the speech signal to obtain initial spectral features; inputting the initial spectral features into a pre-trained speech separation module for background music suppression to output target spectral features; and inputting the target spectral features and the initial spectral features into a pre-trained speech recognition module to output recognized text; wherein the speech separation module comprises an explicit music harmonic convolutional encoder, an implicit music perception encoder and a gated attention fusion processor, and the speech recognition module comprises a noise perception attention mechanism and a convolutional self-attention hybrid encoder.

The invention proposes a method and a device for arithmetic encoding of a current spectral coefficient using preceding spectral coefficients. Said preceding spectral coefficients are already encoded and both, said preceding and current spectral coefficients, are comprised in one or more quantized spectra resulting from quantizing time-frequency-transform of video, audio or speech signal sample values.Said method comprises processing the preceding spectral coefficients, using the processed preceding spectral coefficients for determining a context class being one of at least two different context classes, using the determined context class and a mapping from the at least two different context classes to at least two different probability density functions for determining the probability density function, and arithmetic encoding the current spectral coefficient based on the determined probability density function wherein processing the preceding spectral coefficients comprises non-uniformly quantizing absolutes of the preceding spectral coefficients for use in determining of the context class.

The present application relates to the field of communications, and provides a communication method. The method comprises: sending first capability information, the first capability information being used for indicating the maximum length of a pilot sequence supported by a first communication device during time-frequency transform of the pilot sequence; and receiving a first time-domain signal by means of a first channel, the first time-domain signal carrying a first pilot sequence, the length of the first pilot sequence being less than or equal to the maximum length, and the first pilot sequence being used for estimating or measuring the first channel. On the basis of the solution, the complexity of channel estimation can be effectively reduced, the efficiency of channel estimation or measurement can be improved, and the energy consumption for channel estimation can be reduced, thereby facilitating improving the efficiency of charging communication devices on the basis of radio frequency signals.

The invention discloses a two-stage adversarial migration fault diagnosis method based on an F-D index under strong noise variable working condition interference. Comprising the steps of vibration signal acquisition of a source domain and a target domain, time-frequency transformation, construction of a time-frequency convolution feature extraction network (AM-TFCN) with an attention mechanism, and construction of a two-stage adversarial migration model composed of a feature extractor, a label classifier and a domain discriminator and an adversarial training strategy based on an F-D index. According to the method, supervised training can be carried out on a feature extractor and a label classifier by utilizing source domain labeling data, then parameters are migrated, F-D index dynamic adjustment is introduced under the constraint of a gradient inversion layer, the problems of distribution alignment and fault category identification from a non-interference working condition to different noise levels and variable working condition target domains are solved, and the fault classification accuracy is improved. The migration diagnosis precision is remarkably improved, the result fluctuation is reduced, and the two-stage anti-migration fault diagnosis method based on the F-D index is effective.

This invention relates to the fields of smart city and machine learning technology, specifically a method for continuous self-optimization of urban governance models based on data closed-loop feedback. The scheme involves: acquiring raw waveform data and operating condition reference signals from vibration sensors of urban lifeline facilities; generating a time-frequency energy spectrum matrix through time-frequency transformation; decomposing this matrix into slow-changing and fast-changing components using a machine learning model; removing slow-changing components correlated with long-term operating conditions; labeling the remaining slow-changing components as the first component and the fast-changing components as the second component; generating a frequency-related compensation curve based on the difference between the first component and the historical energy spectrum; performing multiplicative correction; and simultaneously calculating a coupling health index, triggering a calibration work order and updating the historical energy spectrum when the index falls below a threshold. This invention achieves closed-loop optimization of sensor coupling state self-sensing, self-compensation, and self-calibration, significantly improving the long-term data stability and accuracy of urban lifeline monitoring systems.

The invention discloses a harmonic texture and sequence reasoning fused audio-to-MIDI method and system, and relates to the technical field of audio processing. Aiming at the problem of insufficient transcription accuracy of the existing polyphonic music, the method comprises the following steps: carrying out time-frequency transformation on the polyphonic music to generate a spectrogram; constructing a time-frequency target detection network, detecting notes based on harmonic comb texture features, and outputting a time-frequency bounding box; constructing a sequence inference network to extract time sequence context features; carrying out local alignment and weighted fusion on the time sequence context features by using a time-frequency bounding box by adopting an ROI-guided gating fusion mechanism; and inputting the fusion features into a decoder to generate an MIDI sequence, and correcting the note time value by using a time-frequency bounding box. According to the method, the visual texture features and the music sequence logic are effectively fused, and the time precision and robustness of polyphonic music transcription are remarkably improved.

This application discloses a bedridden patient exit warning and vital sign monitoring system based on millimeter-wave radar. The millimeter-wave radar front-end is responsible for emitting electromagnetic waves and collecting the original intermediate-frequency echo signal of the bedridden area; the signal processing module performs time-frequency transformation and clutter filtering on the original signal; the bedding feature identification module senses the bedding status of the clutter-filtered signal and outputs compensation parameters; the vital sign adaptive extraction module restores the signal based on the compensation parameters and extracts vital signs; the exit intention prediction module integrates point cloud data and vital signs to extract multi-dimensional features, predicts the exit intention through an artificial intelligence model, and triggers warnings in stages. Belonging to the field of millimeter-wave radar early warning and monitoring technology, this structure not only achieves high-precision vital sign monitoring in all seasons and under all working conditions, but also issues early warnings as early as when the patient sits up and leans forward.

The application discloses a kind of multi-axis synchronous control methods based on VME and Transformer, comprising: collecting multi-source shafting dynamic data, pre-processing generates standardized multi-source shafting dynamic dataset;Synchronization error signal of main shaft and slave shaft is constructed, and synchronization error time series is formed;Time-frequency transform is carried out using Wigner-Ville distribution algorithm, and multi-axis joint time-frequency feature sequence is constructed;Improved Longformer model is input, and synchronization error prediction result and multi-axis correlation characteristics are generated;Multi-axis synchronous control instruction is generated, and control instruction is issued through VME bus;Feedback signal is received to form closed-loop correction information, and model parameters are updated to realize real-time closed-loop control.The application realizes high-precision synchronization error prediction and real-time synchronous control of semiconductor equipment multi-axis system under high speed and high disturbance condition by introducing improved Longformer model and Wigner-Ville.

This disclosure relates to the field of insulating material testing, including a method and system for evaluating the nonlinear strength of insulating materials based on high-frequency component analysis. The method involves applying an excitation electrical signal of a first preset frequency to the insulating material under test; acquiring the response signal obtained by the insulating material in response to the excitation electrical signal based on a second preset frequency; performing time-frequency transformation on the response signal to obtain the frequency characteristics of various preset higher harmonic components; and determining the evaluation result of the nonlinear strength of the insulating material based on the frequency characteristics of these higher harmonic components. This method allows for obtaining the evaluation result of the nonlinear strength with a single excitation and response acquisition; it eliminates the need for multi-voltage point step testing in traditional techniques, shortening the testing cycle, improving testing efficiency, and enabling efficient, accurate, and standardized determination of the nonlinear strength evaluation result.

The invention discloses an alternating current filter defect detection method, which relates to the technical field of image analysis, and comprises the following steps: constructing a one-dimensional current signal into a time-frequency image through time-frequency transformation, and introducing a visual model containing a self-attention mechanism to carry out global representation learning. Superposition features of concurrent defects can be naturally separated in a time-frequency two-dimensional domain, and compared with a traditional method only depending on a time domain or frequency domain rule, feature decoupling and difference amplification are easier to achieve; defect types, defect occurrence time periods and phase / branch serial numbers are output at the same time through a multi-task framework, so that classification and positioning are subjected to collaborative convergence in a unified network, and error accumulation caused by first classification and then positioning or multi-model series connection is avoided; multi-channel input is formed by three phases and unbalance, utilization of inter-phase correlation and asymmetric clues is enhanced, source phases can be distinguished, and abrupt change direction confusion is reduced.

The present invention relates to a system and method for real-time screening of Attention Deficit Hyperactivity Disorder using electroencephalographic signals processed through a GPU-accelerated time-frequency inference architecture. The invention enables continuous acquisition of multi-channel electroencephalographic data, adaptive preprocessing for artifact suppression and signal stabilization, and parallel execution of time-frequency transformations to extract neurologically relevant features in real time. Extracted features are analyzed using an inference process configured to identify neurological patterns associated with Attention Deficit Hyperactivity Disorder, while continuous validation of signal quality and temporal consistency ensures diagnostic reliability. The system dynamically adapts analytical parameters based on evolving signal characteristics and regulates computational workload to achieve energy-efficient operation during prolonged monitoring. The invention provides a technically integrated, machine-implemented screening solution capable of delivering reliable, real-time neurological assessment suitable for clinical environments.

The application discloses an air gesture recognition method and device, equipment and storage medium, and relates to the technical field of gesture recognition, which comprises the following steps: collecting high-frequency acoustic signals generated by a user performing a gesture in the air through a microphone; performing time-frequency conversion on the high-frequency acoustic signals to generate a two-dimensional time-frequency spectrum; performing spectrum clipping on the two-dimensional time-frequency spectrum to retain target spectrum information in a preset high-frequency band; filtering the target spectrum information to extract potential gesture segments; inputting the potential gesture segments into a gesture recognition model to output a gesture category, and triggering a corresponding interaction instruction according to the gesture category. The application can improve the high-precision recognition of gestures.