259 results about "Signal classification" patented technology

The invention relates to the field of bionic hidden underwater acoustic communication, in particular to an underwater acoustic communication hidden performance evaluation method, device and equipment based on a neural network, and the method comprises the steps: carrying out the feature extraction of a to-be-evaluated underwater acoustic signal, and obtaining a feature sequence of the to-be-evaluated underwater acoustic signal; converting the feature sequence into a two-dimensional matrix with a preset size; the two-dimensional matrix is predicted according to a pre-trained underwater acoustic signal classification neural network model, a probability matrix is obtained, the probability matrix comprises a first probability and a second probability, the first probability represents the probability that the to-be-evaluated underwater acoustic signal belongs to the bionic modulation signal, and the second probability represents the probability that the to-be-evaluated underwater acoustic signal belongs to the bionic modulation signal; the second probability represents the probability that the underwater acoustic signal to be evaluated belongs to the real marine organism signal, and the underwater acoustic signal classification neural network model comprises an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a global pooling layer and a full-connection output layer which are connected in sequence. The dependence on artificial feature design in a traditional recognition method is reduced, and the recognition stability in a complex underwater environment is improved.

The invention relates to an electroencephalogram signal classification method based on multi-scale causal convolution and KAN attention. The electroencephalogram signal classification method comprises the steps of 1, EEG data collection and preprocessing; 2, EEG data enhancement; and 3, motor imagery task classification based on EEG signals. According to the method, the spatial-temporal characteristics of the EEG signals under different frequencies can be effectively extracted, the multi-scale nonlinear characteristics are efficiently fused and weighted, and the problem that nonlinear fitting of the EEG signals in a motor imagery task is insufficient is solved, so that more accurate and reliable motor imagery classification is realized.

The invention relates to the technical field of electromagnetic compatibility and signal processing, and discloses a multi-dimensional feature fusion-based electromagnetic leakage signal classification and identification method, which comprises the following steps of: preprocessing an original electromagnetic signal to obtain standardized data, extracting time domain, frequency domain and space domain features to construct a nine-dimensional feature vector, and extracting a three-dimensional feature vector; constructing an initial feature library and completing the initial feature library through a self-supervision verification mechanism, training the initial feature library through a CNN-LSTM fusion model with an attention mechanism to obtain a classification model, classifying signals to be identified, starting a self-supervision reinforcement learning mechanism optimization model according to an F1 score, and dynamically optimizing the feature library each month; the method solves the defects of the traditional technology, improves the recognition accuracy and the system adaptability, and can be used for cable information safety leakage protection.

The invention discloses a phi-OTDR (Optical Time Domain Reflectometer) vibration signal classification method based on multi-scale residual convolution and a bidirectional long-short-term memory network, relates to the technical field of phase-sensitive optical time domain reflectometer vibration sensing, and solves the problems of insufficient feature extraction, single model fusion mechanism, poor system real-time performance and the like of an existing phi-OTDR vibration signal classification method. The method is realized through the steps of a multi-scale deep convolutional network, a bidirectional long-short term memory network, a cross attention fusion module, a learnable gating mechanism and the like. A cross attention fusion module is adopted, association between spatial features and time sequence features is dynamically established in a double-path network, feature interaction and information coupling are deepened, and the sensing ability of the model to key information is remarkably improved. Meanwhile, the fusion proportion of the multi-path features is adaptively adjusted according to the actual distribution of the input features, background noise interference is effectively suppressed, and the robustness and discrimination performance of the model are enhanced. Accurate identification and classification of external vibration signals are realized.

The invention discloses a Conformer-based multi-task wireless communication signal classification method, and belongs to the technical field of crossing of signal processing and artificial intelligence. In order to solve the problems of task modeling isolation, insufficient feature sharing mechanism and weak model generalization ability in existing wireless communication signal classification, the method comprises the following steps: acquiring an IQ signal, performing dimension raising through a front-end convolution module, inputting the IQ signal into a Conformer network fusing local convolution and a global attention mechanism, and performing depth time sequence feature extraction; and synchronously realizing discrimination of a signal-to-noise ratio grade, a channel type, a modulation mode and a communication system by using a parallel multi-task classification head. The method can improve the classification accuracy and the model generalization ability, reduces the consumption of computing resources, and is suitable for signal recognition in wireless communication, radar and Internet of Things systems.

A radio frequency (RF) signal classification device may include an RF receiver configured to receive RF signals, a quantum computing circuit configured to perform quantum subset summing, and a processor. The processor may be configured to generate a game theory reward matrix for a plurality of different deep learning models, cooperate with the quantum computing circuit to perform quantum subset summing of the game theory reward matrix, select a deep learning model from the plurality thereof based upon the quantum subset summing of the game theory reward matrix, and process the RF signals using the selected deep learning model for RF signal classification.

The invention provides an AI-driven multi-wireless-signal candid photographing detection device which comprises a mounting shell, a hardware acquisition layer, an edge computing layer, a cloud federation layer and a man-machine interaction layer, and all the layers achieve signal interaction through PCIe4.0 and SPI hardware interface communication protocols. The broadband patch antenna, the ceramic antenna and the helical antenna are adopted to cover a full target frequency band, the millimeter wave radar chip is matched, the 16 * 16 metamaterial RIS panel is additionally configured, and the weak signal receiving power is improved by synchronously controlling a reflection phase focusing signal, so that the problem of missing detection is solved; the multi-signal tensor decomposition unit and the CNN-LSTM signal classification unit construct frequency, intensity, time and phase four-order tensors, and overlapped signals are separated through Khatri-Rao space-time coding; a three-layer CNN and a two-layer LSTM are combined, precise classification of candid photographing signals is realized, the misjudgment rate is reduced, millimeter wave assisted precise distance measurement is combined with millimeter wave flight time data and an RSSI attenuation model, a distance measurement error is controlled through a dynamic weighting algorithm, and the problem of positioning misalignment is solved.

The invention provides a desktop card refreshing method, device and equipment for a vehicle-mounted terminal, and a medium, and effectively solves the problem that an existing refreshing mechanism on the vehicle-mounted terminal cannot timely and effectively refresh on the basis of ensuring the performance of the vehicle-mounted terminal. The method comprises the steps that a signal analysis module obtains various vehicle body signals generated on a vehicle through a CAN bus, and a pre-established card mapping relation is inquired based on the various vehicle body signals through a signal classifier to obtain corresponding signal types; classifying and registering various vehicle body signals to a pre-configured corresponding card refreshing channel based on the signal types; the vehicle body signal is refreshed through a target refreshing mechanism in a card refreshing channel, a rendering instruction is generated after the vehicle body signal is processed, and the rendering instruction is sent to a desktop rendering module; and in response to the rendering instruction, controlling the desktop rendering module to execute the rendering instruction so as to refresh a target card picture corresponding to the rendered vehicle body signal on a display screen.

The invention provides a 2D-MUSIC beam line damage positioning method and system, and relates to the technical field of material safety monitoring. By collecting reference guided wave signals at different temperatures, amplitude-phase changes of direct waves of all sensor channels are analyzed, and a temperature-induced amplitude-phase error matrix is constructed; the error matrix is introduced into a two-dimensional multiple signal classification (2D-MUSIC) algorithm, an ideal steering vector is corrected, and a spatial spectrum under an actual propagation model is reconstructed; a cost function is established in combination with noise subspace characteristics, and high-precision and high-robustness positioning of tiny damage is realized through iterative optimization and synchronous inversion of damage positions and channel error parameters. According to the method, the applicability and the positioning accuracy of the 2D-MUSIC algorithm in a variable-temperature environment are remarkably improved.

The invention discloses a hand motion FNIRS signal classification method based on t distribution optimization feature extraction, and the method specifically comprises the steps: employing a portable functional near-infrared collection device to collect an optical density signal of a forehead cortex brain region during the motion execution period during the operation of the device; after pretreatment, the concentration of oxyhemoglobin HBO and the concentration of deoxidized hemoglobin HBR are calculated according to the corrected Beer-Lambert law; according to motion execution and resting state segmentation data, time domain features and time-frequency domain features of oxyhemoglobin concentration data signals of each channel of each experiment test are extracted, and a first feature matrix is formed; performing kernel density estimation and kurtosis and skewness test on the first feature matrix to obtain a second feature matrix, performing t distribution on each corresponding feature satisfying normal distribution in the second feature matrix of the motion execution and resting states, and optimizing feature selection on the basis to obtain an optimized third feature matrix; the optimized third feature matrix is used as classifier input, a brain-computer interface system model training result is obtained, and effective feature selection and extraction are achieved; according to the method, the brain-computer interface system feature extraction of the hand motion FNIRS signals can be optimized, so that the classification performance is improved.

The invention discloses a partial discharge signal identification method and device based on multi-dimensional features and a medium, and the method comprises the steps: collecting a partial discharge signal and a random interference pulse generated in the operation process of a transformer substation, and constructing an original data set according to the partial discharge signal and the random interference pulse; expanding the original data set through a data enhancement technology to obtain a sample data set after sample expansion; extracting a multi-dimensional feature corresponding to each signal sample from the sample data set; wherein the multi-dimensional features comprise a time domain feature, a frequency domain feature and a time-frequency domain feature; and based on the multi-dimensional features, training through a support vector machine technology to obtain a signal classification model so as to classify the partial discharge signals and the random interference pulses through the signal classification model.

The invention discloses a radar target detection method based on graph node dual-channel feature attention fusion, and belongs to the technical field of radar signal detection, and the method comprises the following steps: 1, carrying out the graph node division of received frame radar echo data; 2, respectively extracting time domain amplitude and time frequency characteristics from echo time sequence data corresponding to each graph node; step 3, establishing a feature preprocessing sub-network; step 4, constructing a node feature fusion sub-network; 5, constructing a signal classification graph neural sub-network; 6, connecting the feature preprocessing sub-network, the node feature fusion sub-network and the signal classification graph neural sub-network in series to form a radar target detection neural network; and 7, inputting the test set into the trained radar target detection neural network, and outputting a dichotomy result of which the corresponding node is a target or clutter signal. Through the scheme, the target detection capability of the radar in the clutter environment can be improved.

The invention provides a LoRa positioning method and system based on symbol-level frequency hopping, and the method comprises the steps: a commercial LoRa node generates a LoRa signal, and transmits a signal containing symbol-level frequency hopping through dynamically adjusting the carrier frequency; the receiving gateway receives the symbol-level frequency hopping signal, aligns a demodulation window by using a time-frequency window alignment algorithm and eliminates frequency offset; and the operation server integrates cross-band channel state information of the frequency hopping signals through coherent multi-channel splicing, calculates a power delay spectrum, jointly estimates an angle of arrival and flight time by using a multi-signal classification algorithm, and carries out fusion calculation to obtain a target node position. According to the method, the positioning precision is remarkably improved, and the positioning delay is greatly shortened.

The invention relates to a major network alarm leakage prevention system, method and device based on event research and judgment, and a medium, and the system comprises a multi-level alarm signal classification module which is used for carrying out the classification and labeling management of alarm signals; the main-auxiliary combined intelligent monitoring module is used for integrating warning signals of primary equipment and an auxiliary control system and performing intelligent filtering; the event analysis engine is used for realizing event analysis of the alarm signal based on the knowledge base and the topological relation; the intelligent missing monitoring prevention mechanism module is used for preventing important signal missing monitoring through event grading and intelligent patrol; the intelligent notification and closed-loop management module is used for realizing whole-process management from event generation to filing; the method can effectively prevent the important alarm signals of the main network from missing monitoring, and solves the problems of excessive signals, difficult analysis and easy missing monitoring in a traditional manual monitoring mode.

The invention discloses an electroencephalogram signal classification method, device and equipment and a medium, and is applied to the field of signal processing.The electroencephalogram signal classification method comprises the steps that nonlinear features of target electroencephalogram signals are mapped to a two-dimensional plane, and a recurrence plot of the target electroencephalogram signals is obtained; performing recursive quantitative analysis on the recursive plot to obtain quantitative data of the target electroencephalogram signal; the target electroencephalogram signals are classified by utilizing a pre-trained electroencephalogram signal classification model in combination with the quantized data, a classification result of the target electroencephalogram signals is obtained, the electroencephalogram signal classification model is constructed based on a resistive random access memory array, and the resistive random access memory array comprises a first column used for storing positive weights and a second column used for storing negative weights. According to the method, the nonlinear features in the target electroencephalogram signals can be fully combined, the classification accuracy and classification efficiency of the target electroencephalogram signals can be improved, and the electroencephalogram signals of the epileptic in different states can be accurately classified from a large number of electroencephalogram signals.

A motor imagery electroencephalogram signal classification method based on multi-head self-attention mechanism and parallel convolution belongs to the field of computer software. In view of the problem that the low signal-to-noise ratio of the electroencephalogram signal leads to difficult feature extraction, an improved network model based on EEGNet is proposed, which is referred to as EEG-MATCNet. First, the original electroencephalogram signal is subjected to preliminary feature extraction by using a parallel convolution layer, and different scale convolution kernels can extract time features of different time steps. At the same time, the attention weights of the electroencephalogram signals between the electrodes are calculated by using a multi-head self-attention mechanism, so that the network can better extract spatial features during training. In addition, the receptive field of the convolution kernel is improved by using a time convolution network, so that the model can extract higher-level time features. Experiments prove that the classification method proposed in the application can more effectively improve the feature extraction and classification performance of the motor imagery electroencephalogram signal.

The invention discloses a motor imagery electroencephalogram signal classification method and device, a terminal and a storage medium, and relates to the field of biomedical signal processing.The method comprises the steps that electroencephalogram signals based on user motor imagery are obtained and preprocessed, and electroencephalogram features are determined; performing multi-scale spatio-temporal feature extraction and space and channel decoupling reconstruction on the electroencephalogram features, and determining target spatio-temporal enhancement features; and classifying the target space-time enhancement features through a classification output layer, and determining a classification result. Due to the fact that space and channel decoupling reconstruction is carried out on the features, redundant correlation of the cross-electrode electroencephalogram signals is systematically eliminated, and the problems that in the prior art, space and channel information are jointly processed, information redundancy is caused, and calculation burden is increased can be effectively solved.

This invention relates to a method for estimating the azimuth angle of high-frequency over-the-horizon radar signals based on an optimized layout of multiple orthogonal dual magnetic antennas. The method includes: constructing a compact optimized layout adapted to a shipborne platform using two centrally co-located and orthogonally placed magnetic antennas as basic units; and proposing an azimuth angle estimation method based on this layout: first, using information from the ship's Automatic Identification System (AIS) to perform amplitude and phase calibration on the received signal; then, normalizing the signal's magnitude; finally, using an ideal steering vector and a multi-signal classification algorithm to estimate the direction of arrival (DOA) of the target signal. This method inherits the target detection performance advantages of magnetic antennas while effectively adapting to the spatial constraints of shipborne platforms. The proposed azimuth angle estimation method effectively eliminates the influence of environmental interference on signal reception, achieving azimuth estimation for weak target signals, and providing key technical support for the practical application of small shipborne high-frequency over-the-horizon radar.

The invention provides a lightweight model construction method for micro-seismic signal classification, and belongs to the technical field of micro-seismic signal analysis, and the method comprises the steps: obtaining a micro-fracture signal collected by a micro-seismic monitoring system, and carrying out the preprocessing of the micro-fracture signal, and obtaining a micro-seismic signal classification data set; the method comprises the following steps: constructing an initial lightweight microseismic signal classification model which comprises an input layer, a lightweight feature extraction backbone network and a classifier which are connected in sequence, the lightweight feature extraction backbone network comprises a down-sampling unit, three feature extraction units and a cavity convolution pyramid pooling unit which are connected in sequence; each feature extraction unit is formed by alternately stacking an enhanced shuffling network unit module and a double-attention adaptive residual shrinkage module; performing model training iteration on the initial lightweight micro-seismic signal classification model by using the training set until convergence; verifying the trained lightweight micro-seismic signal classification model by using the verification set, and adjusting and re-training model parameters; and testing the lightweight micro-seismic signal classification model by using the test set, and counting evaluation indexes.

Provided in the present application are an electroencephalogram signal classification method and apparatus, and a device. The method comprises: a first resistive random access memory in a resistive random access memory array acquiring an electroencephalogram signal to be classified, and extracting a signal feature of said electroencephalogram signal; when the similarity between the signal feature and a preset normal electroencephalogram signal feature is not greater than a preset threshold value, calling a target operation program and a target weight value that corresponds to each signal abnormality type to process said electroencephalogram signal, so as to obtain an electric-current feature value corresponding to each target weight value; and determining a signal abnormality type corresponding to the largest electric-current feature value among all electric-current feature values to be a target signal abnormality type of said electroencephalogram signal, wherein the target weight value is the difference between a first weight value and a second weight value that correspond to a signal abnormality type. By means of the present application, the data processing volume and power consumption of a neural network are effectively reduced, and the classification efficiency is improved.

The invention provides a similarity distribution-based pseudo label construction method and a correlation contrast learning method, pseudo labels are generated for instances by calculating similarity distribution among the instances, and the pseudo labels not only provide additional supervision signals for the model, but also can be used as the training is carried out. The change of similarity distribution can also be used for dynamically adjusting the label weight, so that the generalization ability of the model is improved; secondly, the similarity distribution among the instances is constructed, and the mutual relation among different views is also calculated; in multi-view learning, data may be observed from multiple angles or feature spaces, each view may contain different information about the data. By constructing similarity distribution between views, the relationship between instances can be better modeled, so that when one view of the same sample is similar to or dissimilar from other samples, the other view can keep the relationship as well.

This invention relates to a modulation and coding scheme and a method, system, device, and medium for predicting spatial stream counts. The method includes: acquiring received signal strength indication data and preset threshold information between at least two access points and a site; determining the synchronous / asynchronous communication status between each access point based on the dynamic relationship between the received signal strength indication data and the preset threshold information; classifying signals transmitted from adjacent access points to the site as interference signals and signals transmitted from adjacent sites as ambient noise based on the synchronous / asynchronous communication status; calculating the signal-to-noise ratio (SNR) of the site based on the signal classification results; inputting the SNR, synchronous / asynchronous communication status, and access point transmit power into a gradient boosting decision tree model; and outputting the modulation and coding scheme and spatial stream count prediction results of the target access point through the gradient boosting decision tree model. This method significantly improves the MCS / NSS prediction accuracy of high-density WLANs.

The embodiment of the application discloses a method and equipment for reinforcing training of a Gaussian classification neural network, the method comprising: reinforcing training of an original Gaussian classification neural network to obtain a target Gaussian classification neural network; acquiring an image signal to be classified; inputting the image signal to be classified into the target Gaussian classification neural network for processing to obtain a classification result; wherein the reinforcement training neural network comprises a reinforcement network loss layer connected with a convolution-full connection layer, which replaces a Gaussian mixture discriminant layer in the original Gaussian classification neural network to train the Gaussian classification neural network, improves the classification performance of the Gaussian classification neural network, and can realize special application requirements such as feature coding, and when the image signal to be classified is input into the target Gaussian classification neural network, the accuracy of signal classification prediction can be improved.

The application provides a LoRa positioning method and system based on symbol level frequency hopping, and the method comprises the following steps: a commercial LoRa node generates a LoRa signal, and sends a symbol level frequency hopping signal by dynamically adjusting a carrier frequency; a receiving gateway receives the symbol level frequency hopping signal, aligns a demodulation window by using a time-frequency window alignment algorithm, and eliminates frequency offset; a calculation server integrates cross-band channel state information of the frequency hopping signal by coherent multi-channel splicing, calculates a power delay spectrum, jointly estimates an angle of arrival and a time of flight by using a multiple signal classification algorithm, and obtains a target node position by fusion calculation. The method of the application significantly improves positioning accuracy and greatly shortens positioning delay.

The application belongs to the technical field of electric digital data processing, and particularly relates to an adaptive ensemble learning model optimization method for electroencephalogram signal classification, which comprises the following steps: acquiring a multi-channel electroencephalogram sample segment, extracting features containing time-frequency energy, phase locking and spatial mode, the phase locking containing phase locking values among the multi-channels, and constructing three base learners; constructing a weighted brain network through the phase locking values and constructing a brain state feature vector; taking the fusion weight of each base learner as an optimization variable, combining the weight into a particle, grouping according to feature preferences, constructing fitness, iteratively updating the particle based on the fitness, and iteratively optimizing to obtain an optimal weight; weighting and fusing the prediction probabilities of each base learner by using the optimal weight, selecting the category corresponding to the maximum prediction probability as the classification result, and triggering re-optimization based on the change of the adjacent brain state feature vector. The method realizes online adaptive optimization of the weight, and improves the accuracy and long-time stability of motor imagery electroencephalogram signal classification.

The application discloses a motor imagery electroencephalogram signal classification method based on parallel DAMSCN-LSTM, which comprises the following steps: S1, pre-processing four types of motor imagery electroencephalogram signals, including removing electrooculogram and electromyogram and performing band-pass filtering; S2, extracting time features of the electroencephalogram signals by using LSTM; S3, extracting time-frequency features of the electroencephalogram signals in different scales by using DAMSCN, and simultaneously introducing spatial attention and channel attention modules; S4, extracting the multi-scale time-frequency features while introducing spatial attention mechanism and channel attention mechanism; S5, splicing the extracted multi-scale time-frequency features and time features, and then realizing feature classification by means of a full connection layer and a SoftMax layer. Finally, the method is verified on a public data set, and compared with related literatures, and the results show that the multi-class motor imagery electroencephalogram signal classification algorithm provided in the application has better classification results.

The application relates to the technical field of disease screening, in particular to an abnormal heart sound detection method based on a space-time attention feature fusion model. The abnormal heart sound detection method first acquires heart sound signal data, pre-processes the heart sound signal data, and obtains pre-processed heart sound signal data; each heart sound signal feature data is determined by performing feature extraction on the pre-processed heart sound signal data; each heart sound signal feature data is subjected to multi-source feature fusion and heart sound classification through a constructed CNN-TCN-Attention network model, and a classification result of the heart sound signal data is obtained. The application directly extracts features from the heart sound signal data, utilizes the CNN-TCN-Attention network model to perform heart sound signal classification, effectively improves the accuracy of abnormal heart sound recognition, and is mainly applied to the field of abnormal heart sound detection.

The invention discloses a composite modulation signal blind identification method based on time-frequency analysis, which comprises the following steps of: firstly, judging and classifying input pulse signals, and then respectively carrying out identification and parameter estimation on determined signal types. According to the scheme, down-conversion analysis processing, signal time-frequency analysis processing and signal classification identification decision processing are firstly carried out on AD sampling intermediate frequency data sent by a digital signal processing board FPGA to obtain an identification result of a radar signal intra-pulse modulation type, and then corresponding parameter estimation is carried out according to the radar signal intra-pulse modulation type. According to two composite modulation styles of frequency coding and linear frequency modulation combination and frequency coding and two-phase coding combination, on the basis of time-frequency parameter characteristics of various composite modulations, related characteristic parameters are accurately extracted, a modulation type identification process is constructed, and time-frequency analysis is performed on radar signals through fast Fourier transform and phase correlation, so that the identification accuracy of the radar signals is improved. And furthermore, intra-pulse parameter estimation is accurately carried out on the radar signal.

This invention discloses a lightweight adaptive optimization method for high-dimensional hyperparameters of convolutional neural networks (CNNs) for non-stationary time-series signal classification. It aims to address the technical challenges of performance degradation in time-series signal classification models and the reliance on expensive real-world evaluations for hyperparameter configuration under non-stationary perturbation scenarios. This method uses a deep convolutional neural network as the core classification carrier, treating the hyperparameter combinations within the deep convolutional neural network as decision variables to be optimized. With robust classification error rate, computational complexity, and training time as core optimization objectives, it constructs a closed-loop collaborative optimization mechanism of "perception-evaluation-decision" and utilizes a meta-learning dual-branch convolutional polynomial surrogate-assisted evolutionary algorithm (MetaDCP-SAEA) to achieve efficient configuration. This method requires no manual intervention; the convolutional neural network used for classifying non-stationary time-series signals can automatically search for the optimal hyperparameter combination. In simulated non-stationary noise environments, the reduction in classification accuracy can be controlled within 9.17%. It is suitable for robust classification scenarios of non-stationary time-series signals such as industrial IoT monitoring and medical signal diagnosis. It helps to lower the engineering threshold of artificial intelligence technology, promotes the large-scale application of automatic machine learning in complex environments, and has broad market prospects and application value.

The present application relates to a kind of electroencephalogram classification method based on physical information residual polynomial network, design to be trained network, first with polynomial feature extraction layer carries out nonlinear feature extraction, then through LSTM network capture long-term time series dependence of electroencephalogram, then realize the neural network architecture of E / I path separation, respectively to interface excitatory pathway and inhibitory pathway, then by the feature correlation analysis layer of embedding Wilson-Cowan neural population dynamics equation carries out physical constraint, finally in succession fusion layer, classification layer, classification layer completes the construction of to-be-trained network, then based on each sample formed by each multi-channel electroencephalogram, for to-be-trained network training, obtains electroencephalogram classification model, while guaranteeing electroencephalogram classification prediction accuracy, significantly enhance biological explainability.