156 results about "Frequency resolution" patented technology

The invention relates to the technical field of fault diagnosis, in particular to a wind driven generator fault diagnosis method and system based on Mamba-ResNet. The method comprises the following steps: carrying out feature extraction and feature fusion by utilizing preprocessed data, namely constructing adaptive window short-time Fourier transform (AW-STFT) to carry out dynamic time-frequency resolution analysis, carrying out parallel feature extraction and constructing a multi-dimensional heterogeneous feature vector, and carrying out a cross-modal adaptive gating fusion mechanism based on a bidirectional cross gating unit; the method comprises the following steps: constructing a Mamba-ResNet hybrid deep network model architecture; performing model training on the constructed network model architecture; and performing fault diagnosis on the wind driven generator by using the trained model architecture. A tedious manual feature design process in a traditional method is avoided, and the automation level and adaptability of a diagnosis system are remarkably improved.

The present invention relates to the technical field of signal processing and fault diagnosis. The method of the present invention comprises: collecting a flame chemiluminescence signal, and performing adaptive filtering noise reduction processing; using an improved proper orthogonal decomposition method and a high-order mode to obtain comprehensive combustion feature information; developing a real-time synchrosqueezing transform algorithm to perform time-frequency analysis on data, and updating energy distribution in real time; taking into account sensor data to perform multi-parameter joint analysis to obtain a combustion state evaluation, using a plurality of high-speed cameras to acquire flame images from different angles, reconstructing three-dimensional flame morphology by means of a computer vision technology, and providing fault detection information; and constructing a fault prediction model to give an early warning of a combustion fault. The present method implements nondestructive and rapid detection of oscillating combustion faults, keeps rich information of flame images, reflects overall flame pulsation characteristics, increases the speed and accuracy of detection, and compensates for the defect of low frequency resolution of conventional time-frequency analysis methods.

The invention provides an inhaul cable self-vibration fundamental frequency automatic identification method and system based on a harmonic product spectrum, and relates to the technical field of structure monitoring, and the method comprises the steps: receiving vibration acceleration data of an inhaul cable, carrying out the windowing processing of the vibration acceleration data through employing a Hanning window, and generating a windowed signal; carrying out fast Fourier transform on the windowed signal, and improving the frequency resolution by implementing zero filling to obtain an initial frequency spectrum; smoothing the initial frequency spectrum to form a smooth frequency spectrum; performing passband clipping on the smooth frequency spectrum to zero the passband outer amplitude to form a clipped frequency spectrum; calculating the signal-to-noise ratio of the frequency spectrum after cutting, and comparing the signal-to-noise ratio with a preset threshold value; if the signal-to-noise ratio is lower than a preset threshold value, judging that the signal quality is unqualified and terminating the processing flow; and if the signal-to-noise ratio is equal to or higher than the preset threshold value, continuing the normal processing flow. According to the invention, automatic and high-precision identification of the self-vibration fundamental frequency of the inhaul cable can be realized.

According to embodiments, a method performed by a distributed unit (DU) may involve: transmitting, to a radio unit (RU), a first control plane message including section extension information for indicating the number of one or more signal-to-interference-plus-noise-ratio (SINR) values for each physical resource block (PRB) corresponding to a frequency resolution of an SINR report by the RU; and receiving, from the RU, a second control plane message including section type information for indicating the one or more SINR values according to the frequency resolution and the number of the one or more SINR values for each PRB.

The invention discloses a bearing fault diagnosis method based on wavelet time-frequency coding and a convolution visual converter. The method comprises the following steps: collecting vibration signals of a bearing under different working conditions; performing continuous wavelet transform on the one-dimensional vibration signal to generate a two-dimensional time-frequency diagram with high time-frequency resolution; carrying out coding and image enhancement on the time-frequency graph through an enhanced color mapping mode; inputting the time-frequency graph into a convolution visual converter model, and realizing local and global combined extraction of vibration features through convolution projection, a self-attention mechanism and a multi-stage feature fusion structure; the features are classified to identify a state of health or type of fault of the bearing. According to the method, the diagnosis precision and robustness in a complex working condition and high-noise environment are effectively improved, and the method is suitable for state monitoring and early fault early warning of various types of rotary mechanical equipment.

The invention discloses a phase modifier bearing fault diagnosis method, which belongs to the technical field of power equipment fault diagnosis, and comprises the following steps of: analyzing independent components, performing blind source separation on a received phase modifier bearing vibration signal, and extracting three types of independent source signals of impact, abrasion and noise; the gradient driving window length is self-adaptive, and the window length of short-time Fourier transform is dynamically adjusted based on the instantaneous frequency gradient of the independent source signal; wavelet packet frequency band energy screening: performing wavelet packet decomposition on the signal after window length adaptive processing, screening a fault characteristic frequency band based on an energy contribution rate, and reconstructing the signal; and enhancing stochastic resonance, and inputting the reconstructed signal into a stochastic resonance system. According to the method, independent component analysis, gradient driving window length self-adaption, wavelet packet frequency band energy screening and stochastic resonance enhanced fourth-order diagnosis chain are constructed, so that multi-stage cooperative processing of phase modifier bearing faults is realized, and the technical problems of time-frequency resolution contradiction, insufficient feature decoupling and weak generalization ability are effectively solved.

The invention discloses a time-frequency analysis method based on local entropy optimization synchronous demodulation reassignment transformation, relates to the technical field of signal processing, and solves the problem that a traditional time-frequency analysis method is not easy to carry out when complex non-stable vibration signals generated by rotating machinery and the like are processed. The problems of difficulty in balancing time and frequency resolution, easiness in energy diffusion, difficulty in multi-component signal separation and the like exist; according to the method, zero padding is carried out on the two ends of a target signal, a discrete Gaussian window is constructed to generate a sliding windowing signal matrix, a projection angle sequence is generated in an angle space, a time axis and a frequency axis are discretized, a demodulation function containing undetermined parameters is constructed to generate a complex number time-frequency representation matrix, the Renyi entropy value of time-frequency sub-blocks is calculated to determine an optimal demodulation parameter, and the optimal demodulation parameter is obtained. And setting a frequency difference threshold to construct a frequency redistribution operator to generate an optimized time-frequency matrix. The method is mainly used for time-frequency analysis of complex non-stationary vibration signals in the fields of rotating machinery and the like, and high-precision frequency modulation feature extraction and enhancement processing are achieved.

The invention relates to the technical field of crossing of medical instruments and artificial intelligence, and provides a hearing threshold detection method based on multi-mode hearing electrophysiological signal fusion. Comprising stimulation frequency otoacoustic emission and auditory brainstem response induced by sweep frequency stimulation sound, distortion product otoacoustic emission induced by dual-tone stimulation and middle ear sound immittance / reflection signals, and a multi-task deep learning model is constructed to perform fusion analysis on the signals. According to the method, accurate quantitative regression prediction of hearing thresholds with any frequency resolution and automatic classification and discrimination of hearing loss properties are synchronously realized, the problems of single evaluation dimension and insufficient complex hearing loss analysis ability in the prior art are effectively solved, the comprehensiveness, accuracy and clinical operation efficiency of objective hearing evaluation are remarkably improved, and the method is suitable for popularization and application. And an innovative system solution is provided for fine diagnosis of the auditory function.

The invention relates to the technical field of wireless carrier communication, in particular to a cluster unmanned aerial vehicle mixed signal separation and identification method, which comprises the following steps of: extracting a receiving strength time sequence from a received cluster unmanned aerial vehicle mixed wireless carrier communication signal, and calculating the sum of absolute values of difference values between continuous points of the sequence under a plurality of time scales, and linear fitting is carried out to establish a mixed channel fractal dimension. According to the method, difference absolute value accumulation and calculation under a plurality of time scales are carried out on a mixed wireless carrier communication signal receiving strength sequence, a mixed channel fractal dimension is established, and the fractal dimension is utilized to realize adaptive setting of an analysis window length and a stepping amount, so that the time-frequency resolution capability of subsequent segmented Fourier transform is improved; after a mixed signal time-frequency spectrogram is generated, a complete discrete time-frequency trajectory set is constructed through multi-frequency unit peak energy detection and association along a time axis, and unmanned aerial vehicle signals with closely overlapped frequencies can be distinguished.

The application discloses a numerically controlled oscillator based on double modulus accumulation and a method for improving frequency resolution, and the numerically controlled oscillator comprises a parameter configuration module, a carry processing module and a main phase accumulator. The parameter configuration module is used for configuring an integer part and a fractional part of a frequency control word based on a target output frequency, and the fractional part comprises a numerator modulus and a denominator modulus. The carry processing module is used for accumulating the fractional part of the frequency control word. On the basis of expanding an arithmetic logic bit width, a parallel pre-judgment architecture is adopted to generate a carry signal to the main phase accumulator according to the numerator modulus and the denominator modulus. The main phase accumulator is used for accumulating the integer part of the frequency control word and simultaneously updating a phase value according to the carry signal, so that a high-resolution frequency output is realized. The application decomposes the frequency control word into the integer part and the fractional part through the double modulus accumulation architecture, dynamically corrects the integer phase by using an accumulation carry mechanism of the fractional part, and realizes a frequency resolution much higher than that of a traditional structure without significantly increasing the bit width of the main phase accumulator.

A method for noise suppression in a hearing instrument includes using an acousto-electric input transducer of the hearing instrument to generate an input signal from ambient sound. A frequency-band-wise noise suppression is applied to a processing signal derived from the input signal. Stationary noise is detected in the respective frequency band, and depending on the detected stationary noise, an amplification factor of the processing signal is set for the relevant frequency band. An analysis adapted to detect stationary and non-stationary noise is applied to the processing signal. The analysis has a lower frequency resolution than the frequency-band-wise noise suppression and, upon detected presence of stationary and / or non-stationary noise in the analysis of the processing signal, the stationary noise in the frequency-band-wise noise suppression is presumed to be detected in each frequency band, and the corresponding amplification factors are set.

A decoder audio device comprises a receiver (101) and an audio signal generator (103). The receiver receives an audio data signal, the audio data signal comprising: a downmix audio signal that is a downmix of a first multi-channel audio signal; an upmix parameter set for a time-frequency segment of the downmix audio signal, the upmix parameter set comprising a level difference parameter, a correlation parameter and a phase difference parameter; and at least one transient parameter indicative of a transient characteristic of the first multi-channel audio signal. The audio signal generator generates an output multi-channel audio signal by upmixing the downmix audio signal according to the upmix parameter and the transient parameter. The decoder audio device uses an artificial neural network (105) having an input node that receives the upmix parameter and the transient parameter. The transient parameter has a different time-frequency resolution than the upmix parameter, and typically has a much roughened frequency resolution. An improved multi-channel audio signal can be generated with little need for additional overhead in terms of processing complexity, resources, and data rate.

The application discloses a high-speed sampling data real-time extraction method based on parallel digital filters, which comprises the following steps: collecting a measured signal through a collection system of a mixed domain oscilloscope to obtain parallel multi-channel signals; realizing large-multiple real-time extraction filtering through a parallel CIC filter and a cascaded parallel half-band filter to obtain serial data; and finally performing serial extraction on the serial data to obtain output data, so that the frequency resolution of the spectrum analysis function of the oscilloscope can be significantly improved without increasing the FFT length.

A wireless communication device includes one or more antenna arrays that receives RF signals across a plurality of frequency bands. The wireless communication device applies band-specific filtering operation to the received RF signals to obtain filtered RF signals. The wireless communication device converts and digitizes the filtered RF signals to obtain digitized intermediate frequency (IF) signals. The wireless communication device executes a plurality of spectrum measurements for the digitized IF signals. Each spectrum measurement of the plurality of spectrum measurements is based on a different offset frequency of a sequence of offset frequencies. Each spectrum measurement is associated with a first frequency resolution. The processor stores phase coherence information for each spectrum measurement. The processor combines the plurality of spectrum measurements based on the stored phase coherence information to generate a spectrum analysis output with a second frequency resolution that exceeds the first frequency resolution.

The invention describes a method for noise reduction in a hearing instrument (1), wherein an acousto-electrical input transducer (E1) of the hearing instrument (1) generates an input signal (E1) from ambient sound (2), wherein a frequency-bandwise noise reduction (21) is applied to a processing signal (V1) derived from the input signal (E1), in which stationary noise in the respective frequency band (Bj, Bk) is detected, and depending on the detected stationary noise, a gain factor (Gk) of the processing signal (V1) for the respective frequency band (Bj, Bk) is set, wherein an analysis (12) is applied to the processing signal (V1), which is configured to detect both stationary and non-stationary noise, wherein said analysis (12) has a lower frequency resolution than the frequency-bandwise noise reduction (21).and wherein, in the case of a detected presence (Y) of stationary and / or non-stationary noise in said analysis (12) of the processing signal (V1), the stationary noise is assumed to be detected in the frequency-bandwise noise suppression (21) in each frequency band (Bj, Bk), and the corresponding gain factors (Gk) are set.

The purpose of the present disclosure is to provide a frequency analysis device capable of determining a frequency with a higher accuracy than a frequency resolution. A frequency analysis device according to the present disclosure divides a signal wave in the time domain into a plurality of sections at predetermined time intervals, and performs discrete Fourier transform on each of the plurality of sections into which the signal wave has been divided. Furthermore, the frequency analysis device calculates a frequency spectrum by combining the signal waves subjected to the discrete Fourier transform in the plurality of sections, and determines a provisionally determined frequency from the frequency spectrum. Furthermore, the frequency analysis device calculates a section phase difference, which is a phase difference between the signal waves in adjacent sections, for at least one set of the plurality of sections. The frequency analysis device corrects the provisionally determined frequency using the section phase difference.

In order to solve the problem of insufficient time-frequency resolution of the existing fractional-order wavelet transform, the invention provides a generalized fractional-order wavelet transform time-frequency analysis method, which comprises the following steps of: converting a fractional-order rotation angle parameter of the fractional-order wavelet transform into a free variable based on a fractional-order wavelet transform theory; according to the generalized fractional order wavelet transform time-frequency analysis method, the time-frequency characteristics of signals are extracted through maximum operation, and therefore the new generalized fractional order wavelet transform time-frequency analysis method can be obtained, and when the method is used for processing multi-component non-stationary signals, the time-frequency resolution higher than that of a traditional time-frequency analysis method (such as short-time Fourier transform and wavelet transform) can be obtained.

Shear wave phase velocity is estimated from measurements of shear wave motion obtained using ultrasound shear wave elastography or other suitable elastography techniques. Shear wave dispersion curves are generated using a combined generalized Stockwell transform and slant wave number-frequency analysis. A modified version of the S-transform is used to control the time-frequency resolution of a time-frequency decomposition of a signal.

The application discloses an intelligent learning question and answer analysis method and application system based on government service, relates to the technical field of artificial intelligence and government service, and comprises the following steps: performing semantic analysis on question and answer data submitted by a user, extracting intention features and entity features to form semantic representation, identifying mutual exclusion relations of cross-department rules in a government rule library, and generating a conflict set. On this basis, the time limit for validity and the time limit for completing the matter of the user are combined to calculate the time limit priority, the conflict resolution path is prioritized, and a multi-path decision tree is constructed. Meanwhile, high-frequency resolution paths are learned from historical question and answer data and a knowledge base is established, so that experience reuse is realized. When a new question and answer request is received, the resolution path in the knowledge base is matched preferentially, and when the matching fails, the optimal reply is output through the decision tree, so that the intelligent level of government question and answer and the cross-department business processing efficiency are improved.

The invention belongs to the technical field of time-frequency spectrum decomposition, and discloses a time-frequency spectrum decomposition method and device based on empirical matching pursuit and a storage medium, the time-frequency spectrum decomposition method comprises the following steps: constructing an empirical mode decomposition (EMD) dictionary suitable for time-frequency analysis according to input seismic data; performing matching pursuit on a seismic signal to be decomposed by using the EMD dictionary to obtain a time-frequency atom combination of the seismic signal; and performing multiple synchronous compression transformation on the time-frequency atoms of the seismic signal to obtain the time-frequency distribution of the seismic signal. According to the method, a new time-frequency characterization method is deduced, a time-frequency spectrum decomposition strategy based on empirical matching pursuit is designed, and the time-frequency resolution of seismic signals is improved by introducing multiple synchronous compression transformation.

A high-concealment signal detection method, system, device and medium based on a digital residual image, the sampled signal IQ data is preprocessed, that is, the key parameters including the observation bandwidth (that is, the sampling rate), the frame length, the overlap rate and the time and frequency resolution are set; at the same time, each frame of signal is subjected to windowing filtering processing, and the system automatically determines the segmentation number according to the signal length; then, each signal segment is subjected to fast Fourier transform, and the calculation result is stored in a bitmap database; then, the frequency spectrum data is continuously superimposed in the bitmap database, based on the total superposition number, the proportion of each cell is calculated to form a probability density distribution matrix, the probability density distribution matrix is subjected to normalization processing, and all probability values are ensured to be located between 0 and 1; by using a self-defined mapping table, the normalized probability density matrix is converted into a digital residual image through linear mapping; finally, by using a method combining Gamma correction and median filtering, the digital residual image obtained through mapping is further enhanced, and a final clear and smooth digital residual image is obtained; in the application, the Gamma correction can effectively enhance the darker part to improve the image contrast, the median filtering can effectively remove the salt and pepper noise generated by image mapping, and the important edges and details of the signal in the image are retained; due to the low calculation complexity, the application is suitable for real-time processing and meets the low complexity requirement.

The application discloses a digital oscilloscope frequency domain trigger method based on a direct digital integration method, a group of cosine waves of specific frequencies are generated through self-defined analysis frequency bands and frequency resolutions, then a measured signal is multiplied by the cosine waves of the specific frequencies and integrated, energy of the signal on the frequency component is directly obtained, and finally the analysis result is used for a frequency domain trigger function in the oscilloscope.

The present invention provides an efficient scheme for improved cross-product high frequency reconstruction (HFR), where a new frequency component of Q Ω + r Ω 0 is generated based on the existing components Ω and Ω + Ω 0.SOLUTION: The present invention performs a block-wise harmonic transposition, wherein time blocks of complex subband samples are processed with a common phase modification. The superposition of the modified samples has the advantageous effect of limiting unwanted intermodulation products, thereby allowing a coarser frequency resolution and / or a lower degree of oversampling to be used. The present invention in one embodiment uses a window function suitable for use with an improved block-based cross product HFR. A hardware embodiment of the present invention includes an analysis filter bank (101), a subband processor (102) configured by control data (104), and a synthesis filter bank (103).SELECTED DRAWING: Figure 1

The invention belongs to the technical field of distributed optical fiber sensing, and relates to an anti-distortion distributed optical fiber sensing method and system based on array phase spectrum demodulation, an ultra-weak reflection optical fiber Bragg grating array is used as a sensing optical fiber, a series of reflection peaks with the signal-to-noise ratio higher than that of RBS are generated, a reflection peak sequence is formed, and the beat frequency signal spectrum amplitude is increased. A discrete beat frequency signal with frequency domain distortion output by an OFDR sensing system is converted to a frequency domain through Fourier transform of a symmetric window function to obtain a corresponding discrete frequency spectrum, and characteristic parameters such as amplitude and phase of a reflection peak sequence in the frequency spectrum are tracked and extracted; and then frequency correction factors of the reflection peak sequence in all measurement periods are calculated by using an amplitude interpolation correction method to obtain a corrected accurate array phase spectrum, and phase difference and strain calculation are performed to realize multi-point quasi-distributed strain demodulation for suppressing frequency domain distortion such as Doppler frequency shift and the like. The frequency resolution limitation caused by the fence effect is broken, and the fidelity of the frequency domain distortion signal is recovered.

The application discloses a segmented adaptive Fourier transform analysis method of power quality disturbance. The method firstly collects power quality disturbance signals in a power grid and performs Fourier transform, acquires peak envelope points in a frequency spectrum through a maximum envelope method to dynamically select characteristic frequency points and calculate frequency spectrum segmentation points; subsequently, a composite objective function is used to optimize Gaussian window parameters for each segment; finally, the signal is analyzed by using the optimized window function to obtain time-frequency results with adaptive resolution. The application allocates higher time resolution to strong time-varying components and higher frequency resolution to near steady-state components through segmented adaptive resolution and a composite objective function optimization strategy, effectively avoids the spectrum overlap and energy diffusion problems that are prone to occur in traditional methods, improves the accuracy and robustness of disturbance feature extraction, and can be widely applied to power grid fault diagnosis, disturbance identification and early warning.

The invention discloses a heart sound signal acquisition and automatic classification method based on deep learning, and aims to solve the problems that wireless transmission time delay of heart sound signals is high, weak pathological features are difficult to extract and classification accuracy is low. The method comprises the following steps: acquiring signals of four auscultation areas by using a high-fidelity wireless digital stethoscope; high-speed and low-delay transmission of medical-level signals is realized by utilizing a star flash transmission module and combining predictive coding and Huffman coding compression mechanisms; in the preprocessing stage, self-adaptive wavelet denoising based on signal energy entropy is adopted, a multi-resolution fusion feature extraction strategy is provided, the problem of time-frequency resolution tradeoff is solved through narrow window and wide window dual-channel parallel processing, and transient positioning features and pathological frequency domain textures are captured at the same time; the classification model adopts a multi-scale parallel convolution structure and is matched with a recurrent neural network and an attention mechanism, so that the problem that transient and continuous heart sound features are difficult to consider at the same time in a single scale is solved. According to the method, the accuracy and robustness of heart sound classification are remarkably improved.

The application discloses a target channel selection output method and system for frequency division multiplexing superconducting detectors, and belongs to the technical field of digital signal processing and superconducting detector readout electronics. The method aims at the problems of high resolution channelization resource occupation, high full spectrum output bandwidth pressure, and obvious spectrum leakage and crosstalk under high multiplexing ratio. On the basis of high resolution channelization processing, the mapping relationship between target frequency points and channels and the target channel selection output mechanism are introduced. After the wideband digital input signal is processed by the polyphase filter bank channelization, only the complex data corresponding to the target frequency points containing effective signals are selected and output, and are further separated into I / Q data stream output. The scheme combines high resolution channelization processing with the mapping driven selection output mechanism, reduces resource occupation and data transmission bandwidth while ensuring frequency resolution, and improves real-time performance and signal fidelity in the readout scene of high multiplexing ratio detectors.

A FMCW radar ranging method, system and medium based on power-weighted centroid, which comprises obtaining the beat signal of the FMCW radar and performing Fourier transform thereon to obtain an initial distance spectrum composed of multiple distance units; performing interpolation processing on at least two distance units corresponding to a preset near distance range in the initial distance spectrum to generate a refined distance spectrum with a higher frequency resolution than the initial distance spectrum; performing power-weighted centroid calculation on a continuous frequency spectrum interval corresponding to the preset near distance range based on the refined distance spectrum to obtain a first distance estimation value; performing peak value detection on the initial distance spectrum to obtain a second distance estimation value; performing time domain filtering processing on the first distance estimation values and the second distance estimation values obtained from continuous multiple frames of signals respectively; obtaining a target distance value based on the smoothed first distance estimation value and the smoothed second distance estimation value according to a preset fusion rule, and outputting.

The utility model relates to the field of radio transceiving, and discloses an ultra-wideband extremely-small-step frequency hopping frequency synthesizer module. According to the invention, the high-speed dot frequency array divides an ultra-wideband frequency 9-18GHz signal into 8 segments, and the 8 segments are respectively shifted and converted to 1000-2200MHz for frequency and phase discrimination, so that the phase discrimination frequency is greatly reduced, excellent phase noise can be obtained, and ultra-wideband frequency coverage is realized at the same time. The adopted high-speed dot frequency phase-locked array is low in phase noise and large in frequency span coverage, the switching time is short through a high-speed switch, large-range fast frequency hopping can be achieved, and powerful support is provided for prolonging the large-range frequency hopping time of a whole circuit system. A DDS direct digital frequency synthesis circuit is adopted, near-end stray is low, high-frequency resolution is provided, and extremely low frequency stepping can be achieved. And the frequency hopping speed is high and reaches below 1uS, so that the rapid frequency hopping of the frequency synthesizer system can be realized.

The invention particularly relates to a radar micro-Doppler spectrum generation method for classification and identification, and the method comprises the steps: receiving an input time domain signal, and carrying out the Fourier transformation of the input time domain signal; rearranging frequency spectrum components obtained by Fourier transform according to an amplitude descending order; setting a spectrum distance threshold value and an intensity threshold value, and gradually eliminating fuselage components through the steps of primary screening, secondary screening and iterative elimination; recovering the frequency spectrum component without the fuselage component to an initial frequency position according to an original frequency index, and executing inverse Fourier transform to obtain a pure time domain signal; and performing time-frequency analysis on the pure time-domain signal by using S transformation to generate a time-frequency matrix, and normalizing an amplitude value of the time-frequency matrix to a [0, 255] gray scale interval to obtain a radar micro-Doppler spectrum. According to the method, a fuselage component elimination method is improved, S transformation with higher time-frequency resolution is selected, a time-frequency atlas data set fully reflecting the micro-Doppler characteristics of the target can be generated, and the method can be used for a convolutional neural network to realize accurate classification and recognition.