Abstract:Low-light image intensifier is a key component in night vision technology. The signal-to-noise ratio is one of the important parameters to evaluate the performance of lowlight image intensifier. A control system of signal to noise ratio tester for image intensifier based on FPGA is designed, including signal acquisition module, constant current driving module, main control module, communication module and power module. The system takes XC7K325T FPGA as the core, complete the ADC, DAC, serial communication and other functions as well as interface logic design, realize the photocurrent data acquisition, signal processing and system parameter setting and other functions, so as to complete the measurement of signal to noise ratio of image intensifier. The test results show that the functions of the image intensifier signal-to-noise ratio tester are normal and the measurement error of the signal-to-noise ratio is less than 2%, which verifies the correctness of the design of the control system and lays the foundation for the reliable application of the signal-to-noise ratio tester.

Abstract:In order to improve the radiation efficiency of the electrically small antenna, a three-layer miniaturized loop antenna based on split resonance ring structure is proposed. The antenna consists of a feeding layer and two coupling layers and is excited by electromagnetic coupling. An E-shaped matching network is designed in the feeding layer to reduce the reverse current, which is the shortcoming of traditional coupling feeder SRR antenna. The coupling ring structure is used to realize strong magnetic coupling and reduce the metal loss. A fixed air gap between the coupling layer and the feeding layer is introduced to effectively reduce the dielectric loss and improve the radiation efficiency. The influence of dielectric substrate on the antenna performances is analyzed by electromagnetic simulation. The proposed antenna is fabricated and measured. The experiment shows that the measured results are basically consistent with the simulated results. The center frequency of the antenna is 50.03 MHz, and the electrical size is 0.045λ0×0.045λ0×0.01λ0 (λ0 is the working wavelength). The antenna has good omnidirectional performance in the horizontal plane, and the radiation efficiency is more than 41%.

Abstract:Bedside care that relies on manual monitoring of sleeping posture is inefficient. To automatically collect accurate sleeping posture information, improve user experience and protect user privacy, a simple, low-cost, high-precision and high-speed sleeping posture monitoring system based on flexible pressure sensing array is designed by combining theoretical analysis and experimental research. Theoretically, according to the similarity between tactile and visual recognition mechanism, the influence of pressure image resolution on the classification effect of sleeping posture is revealed. The 24×24 scale threshold of pressure sensing array is further obtained by analyzing a public dataset, which significantly reduces the cost and complexity of the system. Experimentally, the sleeping posture monitoring system is designed based on 3 modules: Flexible pressure sensing array, signal scanning acquisition circuit and sleeping posture recognition algorithm. Firstly, the piezoresistive sensing array is designed based on flexible pressure-sensitive conductive sheet Velostat. Secondly, the anti-crosstalk signal acquisition circuit is arranged based on the zero-potential array scanning method. Thirdly, the sleeping posture classifier is trained with improved residual network. After experimental testing, the recognition accuracy for 8 types of sleeping posture is 99.57%, and the monitoring rate can reach 150 ms/frame. The system is expected to be commercially used for reliable real-time sleeping posture monitoring.

Abstract:Aiming at the problem of complex factors affecting the objective transfer function of the feedforward controller and the problem of incorrect acoustic test with head and torso simulator in the process of design helmet active noise control headset a method of designing feedforward controller based on human wearing test data is proposed. The principle of active noise control headset is analyzed, which makes up the deficiency of existing theories. Experiments are designed to verify the influence of noise source direction, ear pad elasticity, feedback controller, bandage elasticity, helmet and testing device on the target transfer function of the feedforward controller, some recommendations are proposed. The feedforward controller is designed based on the data of human wearing test, and the effect of hybrid active noise control is tested when it cooperates with the feedback controller. The noise reduction 21.3 dB when the human is wearing, but 10.7 dB when the head and torso simulator is wearing, which proves that the noise reduction effect of headset based on the human wearing test design is better, and the head and torso simulator wearing cannot reflect the human wearing situation very well.

Abstract:For the nonlinear and non-stationary characteristics of rolling bearing signals, reasonable feature selection can improve the fault diagnosis rate. A fault diagnosis model based on multiscale permutation entropy (MPE) and improved whale algorithm (IWOA) was proposed to optimize support vector machine (SVM). Firstly, signal denoising was preprocessed by VMD, and multi-scale permutation entropy was calculated to reconstruct signal features. Secondly, inertial dynamic weights were introduced to improve the whale algorithm, and SVM parameters were trained to establish the IWOA-SVM fault diagnosis model. Finally, the bearing data set of Case Western Reserve University was used for simulation experiments. The results show that, compared with multi-scale entropy, MPE can represent more abundant feature information, and the fault recognition rate is improved by 2.1%. Compared with other optimization algorithms, the fault diagnosis model optimized by IWOA based on SVM has fast convergence speed, short training time and high fault recognition accuracy, which can effectively diagnose rolling bearings.

Abstract:In order to realize the personalized customization and realtime adjustment of the therapeutic parameters of the rehabilitation electrical stimulation system, a closedloop electrical stimulation system for lower limb muscle strength rehabilitation based on modulated medium frequency electrical stimulation was proposed in this paper. A low-frequency modulation and medium-frequency stimulation circuit was designed, and a support vector machine regression prediction model between the electrical stimulation parameters and the angle of the knee joint was established based on the genetic algorithm. A closed-loop feedback system based on fuzzy internal model control PID was built to achieve a more accurate and stable parameter setting effect. The knee motion experiment showed that the subjects were closer to the expected joint motion trajectory without pain. The maximum root mean square error between the knee motion Angle and the expected value in the 30 groups was 10.21°, and the minimum root mean square error was 5.48°. Compared with traditional low-frequency electrical stimulation, the mean amplitude of myoelectric stimulation was increased by more than 20 microvolts. The parameters of the electrical stimulation system proposed in this paper can be realized from person to person, and can be adjusted in real time according to the closed-loop feedback results. The system can effectively activate muscles and improve muscle strength, and has a good application prospect in the gait training of muscle strength rehabilitation.

Abstract:As the information source of intelligent substation, electronic transformer needs realtime monitoring. In order to solve the problems of existing CT monitoring technology in terms of realtime performance, accuracy and storage capacity, this paper designs a set of CT side cloud collaborative realtime monitoring system based on Harmony architecture. The single chip microcomputer is used to collect the transformer data, and the GPS second pulse is used as the reference to ensure the sampling synchronization, and then the Harmony operating system is stored on the cloud, and the Kalman filter and the discrete Fourier algorithm with Cassel window are introduced on the edge computing side to filter the noise and harmonics in the process of data acquisition. When the transformer accuracy is insufficient, the buzzer alarm is triggered. Experimental verification shows that the system realizes real-time, synchronous uploading and mass storage of monitoring data of multiple transformers. The storage capacity depends on the size of cloud server and can be expanded at any time. The system meets the accuracy requirements of transformer realtime monitoring, the amplitude error is less than 0.75%, meets the accuracy level of 0.2, and allows the staff to view the results at any time and in other places. In addition, the edge-cloud collaborative working mode has the characteristics of one-time development and multi-terminal deployment, so that the system has strong portability and can be applied to other different power system scenarios.

Abstract:To achieve accurate temperature measurement in the harsh environment of an explosive temperature field, we designed a temperature data acquisition and storage system. The system uses an FPGA as a microcontroller and armored thermocouples as temperature measurement elements.Its rugged protective structure and stainless steel armor of thermocouple make the system highly survivable in the explosion field. The system has a parameter programmable and multi-triggerable design, making it flexible and reliable. To address the problem of insufficient dynamic response of the thermocouples, we completed the dynamic compensation filter design using particle swarm algorithms and integrated it into the upper computer software. The experimental results show that the system has an average static error of 1.43% and a dynamic error that does not exceed 6.4%.

Abstract:In order to describe the degradation process of key components more accurately, the influence of maintenance activities on the degradation rate and initial degradation of equipment performance was proposed when the Wiener degradation process is used to model equipment. Based on the established model, the probability distribution of the remaining useful life is obtained based on the concept of first arrival time. According to the predicted value of the remaining useful life, the aperiodic performance detection is carried out as the time interval, and the degradation value obtained is compared with the fault threshold and maintenance threshold, the corresponding maintenance activities are made. Then, taking maintenance threshold as variable, Monte Carlo cycle simulation method is used to select the optimal value to achieve the minimum cost-effectiveness ratio in the conditionbased maintenance strategy of key components, so as to reduce the cost-effectiveness ratio during the use of equipment. Finally, the ship component gyroscope is taken as an example, the simulation results show that the relative error of the predicted the remaining useful life value is less than 0.3%. When the maintenance threshold is set as 8.02 degrees/day, the efficiency ratio reaches the lowest as 199.996 yuan/day. The simulation and comparison between periodic detection of conditionbased maintenance strategy show that the proposed method can save the detection cost, cut down the number of the breakdown maintenance, decrease the maintenance cost, and effectively reduce the cost-effectiveness ratio.

Abstract:In food and biological laboratories and other fields, there are strict requirements for the number of planktonic bacteria, which further puts forward higher requirements for planktonic bacteria samplers. In order to solve the problem that the gas flow of the planktonic bacteria sampler are affected by the environment factors, the planktonic bacteria sampler with flow calibration function was researched in this paper, a venturi structure integrating atmospheric pressure and temperature & humidity sensors was designed, the flow compensation formula was improved, a fan control system based on anti saturation integral PID (proportion, integral, differential) was develop, and the real-time compensation and closed-loop control of the gas flow were realized. The VMD(Variational Mode Decomposition) algorithm was applied to filter the measurement data of atmospheric pressure, which reduced the influence of noise on the measurement results. A prototype of planktonic bacteria sampler based on Anderson impact principle was developed, and relevant experimental researches were carried out. The experimental results show that the differential pressure indication error of the instrument is within ±0.5% after calibration, and the flow collection accuracy indication error is within ±1%, which meets the requirements of the flow calibration specification of planktonic bacteria sampler, overcomes the problem of large gas collection flow error caused by environmental factors, and has good application value in the field of planktonic bacteria sampler.

Abstract:Exploring the surface of Mars with a mobile rover is the mainstream way for major spacefaring nations to explore Mars. In order to get a safe path for the rover, a Mars rover path planning algorithm based on the improved A-Star algorithm is proposed. By improving the weight factor in the path information algorithm and optimizing the objective function of the traditional algorithm, a large number of worthless search paths can be omitted, shortening the time of search and improving the search efficiency. Compared with the traditional path planning can shorten 53.94% of the time; the introduction of corner optimization algorithm, in the case of the path length is basically the same, reduce the number of turns in the global path; to meet the needs of efficient and stable operation of the detector, corner optimization, the number of turns can be reduced by 16.77%. The path is smoothed by the fourth-order Bessel curve, effectively avoiding the appearance of corner spikes and ensuring the smooth travel of the rover on the surface of Mars.

Abstract:Using AdaBoost algorithm of boosting learning to solve the suboptimal relays selection can reduce the realtime processing delay and computational complexity in cascaded relaying system, when wireless communication channels are in complex application scenarios such as multi-hop relays and sub-channel assignment. The channel state information of the legitimate channel and the eavesdropping channel is used as the input of the training model, and the index of the relay nodes with a certain value of the security capacity of the system is used as the output to transform the suboptimal relay selection problem of the cascaded relay system into a multiclass classification problem, which is solved by Support Vector Machines based on AdaBoost weighted voting. The suboptimal relay selection scheme for the cascaded relay system can be divided into three phases: generation of dataset, ensemble model training and result prediction. Finally, by plotting the classification accuracy and P-R curves, it is verified that the integrated learning model has higher accuracy for suboptimal relay selection and can improve the performance of relay collaboration.

Abstract:In order to improve the path planning efficiency of unmanned ships when performing water sampling tasks, a path planning algorithm combining ant colony algorithm and firefly algorithm is proposed. Firstly, when constructing the shortest path network, the steering angle cost heuristic function is introduced into the traditional ant colony algorithm to reduce frequent turns in the path search results. Then, redundant nodes in the search results are removed to further reduce the number of turns of the unmanned ship, so that the obtained path is more suitable for the unmanned ship. Finally, when solving the optimal sampling order, an improved firefly algorithm is designed based on random correction, which improves the convergence speed of the algorithm. The simulation results show that the algorithm designed in this paper can complete the path planning task of water sampling task. Compared with the traditional algorithm, the search efficiency is higher and the total path length is effectively shortened.

Abstract:In order to reduce various losses that may be caused by dense fog, the shortterm prediction of dense fog′s trend has become a research hotspot in the field of meteorological shortterm prediction. However, current researches focus on the temporal features of dense fog while ignoring its spatial features, so the prediction accuracy is still at a relatively low level. To this end, this paper proposes an algorithm of short-term prediction for dense fog′s trend based on fusion of spatio-temporal features. The algorithm takes multistations as nodes in the graph data. By advancing Graph Attention Network, we realize the extraction of spatial features. On this basis, combined with time information, adjusting the LongShort Term Memory network to further extract temporal features from the spatial features in order to realize feature-level fusion. Then we use the fully connected layer to output the predicted value of visibility. Further we can get to the predict result t based on the predicted value of visibility. We apply the algorithm on meteorological data released by National Centers for Environmental Information to carry out the experiment. The experimental results show that the F1-score and TS-score of the proposed algorithm take an improvement of 2%~12% on baseline models, which proves that the proposed algorithm has a great practical application value.

Abstract:Aiming at the green heterogeneous fleet vehicle routing problem(GHFVRP), a mixed integer programming (MIP) model with the optimization objective of minimizing the sum of vehicle fixed cost, driving cost and carbon emission cost was established, and an enhanced Lagrange relaxation heuristic algorithm (ELRHA) is proposed to solve it. Firstly, the dual problem is constructed by relaxation difficulty constraint and decomposed into two subproblems. Then the Lagrange multiplier is updated by subgradient method, and the lower bound of the original problem is obtained by solving the two subproblems. Secondly, a two-stage heuristic algorithm is designed to repair and optimize the lower bound to obtain a better feasible solution and update the upper bound. Finally, the simulation experiment is carried out. The experimental results show that 17 examples are tested for 20 times under the same experimental environment. The average solving gap of ELRHA is 4.49%, which is 3.28% higher than Gurobi. Meanwhile, the comparison with other algorithms further verifies that ELRHA can solve the problem with high quality upper bound. It can be seen that ELRHA can effectively solve GHFVRP.

Abstract:For enhancing the visual axis stability accuracy and robustness of the airborne photoelectric stabilization platform, a fuzzy nonsingular fast terminal sliding mode active disturbance rejection control strategy is proposed. Firstly, a nonsingular fast terminal sliding surface is used to enhance the convergence rate of the system state in the sliding phase. Secondly, the fast power reaching law with fuzzy correction term is established to suppress chattering. The introduction of fuzzy logic enables the controller parameters to be adjusted according to the error changes.On this basis, the linear extended state observer is used for strengthening the disturbance rejection function of the system. Then, the stability of the system is analysed by Lyapunov theorem. Finally, a simulation comparison experiment is carried out under the disturbance and parameter perturbation. The simulation results show that the system based on the proposed method is more robust than the other four control algorithms.

Abstract:After the PV power supply is connected to the distribution network, the uncertainty, intermittency and fluctuation of PV power generation increase the difficulty of identifying faults in the distribution network. To address this problem, this paper proposes a method combining the entropy-variance modal component with a neural network to improve the ResNet model. Firstly, a PSCAD simulation model of the distribution network containing PV power is built to obtain batch data under different complex fault scenarios. Secondly, the entropy-variance modal (E-VMD) method is used to reconstruct the feature matrix of the samples, and then the improved residual network is used to further explore the implied features of the fault samples, and then the model is trained and tested. In comparison with the classification results of other models in the literature, the improved ResNet model achieves an average accuracy of 99.95% for fault type identification and 99.75% for fault feeder identification, and has good robustness, which can effectively achieve fast fault identification in distribution networks containing PV power supplies.

Abstract:With the rapid development of science and technology and the sharp increase of data dimensions, it is difficult for traditional dimensionality reduction algorithms to find the optimal subspace of the data, which seriously affects the performance of the classifier. This paper proposes a local ratio sum discriminant analysis based on adaptive subspace graph. The ratio sum discriminant analysis considering the local structure is proposed; the alternative iterative optimization method is used to avoid the suboptimal solution found by the existing ratio sum optimization methods; the nearest neighbor similarity graph is learned in the optimal subspace instead of the original space, so as to avoid it. Influenced by the original spatial noise points; the Shannon entropy constraint is introduced to avoid trivial solutions; finally, the samples are projected to the optimal subspace. On synthetic datasets and face datasets, the proposed algorithm is tested with a large number of SOTA discriminant analysis algorithms for classification tasks. A large number of experimental results show that the proposed algorithm can learn a projection subspace with better discriminant performance and has better classification effect.

Abstract:The variable pitch control of wind turbine is subject to external interference and large parameter changes, resulting in unstable output power, this paper introduces an intelligent control algorithm, adds a fuzzy algorithm on the basis of RBF neural network, and uses fuzzy RBF neural network control to adjust the PID parameters online and in real time.When the actual wind speed is higher than the rated wind speed adjust the blade pitch angle of the fan to change the aerodynamic torque obtained by the fan, make the fan output power stable near the rated power.In this paper, the mathematical model of each module of the fan is established, and the corresponding simulation model is established in the MATLAB/Simulink. According to the experimental results, the control effect based on the above method is better than traditional PID control and conventional RBF neural network PID control, with faster response, less overshoot of wind energy utilization coefficient performance, more stable power output, and more conducive to the system stability of wind turbine.

Abstract:In the traditional transformer winding mechanical fault diagnosis method, only the winding axial vibration is considered, and the feature parameter extraction is complex and the recognition accuracy is low. This paper presents a mechanical fault diagnosis method for transformer windings based on two-axis vibration and multi-sensor fusion. Firstly, the two-axis vibration relationship graph is proposed as the feature image from the perspective of the axial and radial vibration correlation of the winding. Then the lightweight convolutional neural network MobileNet V2 is used to train the image data obtained by different sensors. Finally, the D-S evidence theory is used to fuse the multi-dimensional information source recognition results and make the final decision. The experimental results show that the fault diagnosis accuracy of the proposed method can reach 99.4%. Compared with the traditional fault classification method, the feature extraction step is simplified, and the diagnostic accuracy is improved by more than 6.2%, which provides a feasible scheme for mechanical fault diagnosis of transformer winding.

Abstract:In order to solve the problem of high cost of wafer test, some quality prediction solutions based on spatial correlation have been proposed in the field of adaptive test. But most of these solutions sacrifice too much forecast accuracy in order to reduce costs. To solve this problem, this paper proposes a fine-grained quality prediction method. This method uses the Bad Neighbor Ratio to classify the grains predicted by the spatial correlation model, and selects different test sets for different types of grains. In addition, a spatial verification step is introduced before the selection of the die test set, which can ensure the test quality of the entire solution. The experimental results show that compared with the indirect test method, the average test escape rate of the proposed method is reduced by 83%, and the average test item saving rate is increased by 14%. Compared with the dynamic part average test method, the average test escape rate is reduced by 81%, and the average test item saving rate is increased by 17%.

Abstract:Aiming at the problem of low accuracy of defect detection of coal mine power equipment, this paper proposes a method for defect detection of coal mine power equipment based on an improved YOLOv5s. The method mainly includes three primary modifications. Firstly, a multibranched coordinate attention module is proposed, enhancing the ability of the model to obtain information about defect areas. Secondly, a feature fusion network module is proposed, which further enhances the feature expression and fusion ability of the model by connecting the non-adjacent feature information between the backbone network and the neck network across layers. Finally, a fast spatial pyramid pooling average pooling module is proposed, and the path of the neck network is embedded between the fusion networks to improve the ability of the shallow positioning information of the network to be transmitted to the deep layer. Experimental results demonstrate that the mAP@0.5 of improved YOLOv5s model in increased of 3.1%, and the F1 score is increased by 3%, meeting the detection demands of coal mine power equipment defects and has higher detection accuracy.

Abstract:After the size of the instrument to test the flow resistance is reduced from large to small, its tight arrangement leads to the signal interference of the electronic components and the airflow fluctuation of the equipment is large in the equipment. This phenomenon causes the stability of the test system to deteriorate. Through the diagnosis of interference causes, the simple filtering method makes the test accuracy poor. Through the summary of various problems. Firstly, the median algorithm is used to eliminate the pulse signal. Then the recursive averaging algorithm is used to eliminate the periodic variation of the intake system and the vibration interference of the vacuum pump. Furthermore，the mean absolute error and mean square error are used to define the optimal value of the algorithm. Finally, the data distortion before low-pass filter processing will be reduced. The experimental results show that the test system using the improved median recursion method has better stability, but its response speed is slow. The improved median recursion method can effectively eliminate the interference signals of periodicity, pulse and dither. The static testing system that it applies to has high requirements for testing stability and testing progress, but low requirements for real-time testing.

Abstract:Aiming at the problem that the passive positioning of long-distance targets (such as submarines, torpedoes, etc.) is affected by Marine environment noise and ship's own noise, which leads to the reduction of positioning accuracy, this paper proposes an enhancement method based on convolutional denoising autoencoder and adaptive least mean square error filter (CDAE-LMSAF). By extracting the time-spectral features of the underwater target radiation signal and the noisy signal, it is trained and modeled as the input of convolutional de-noising self-coding. Then, adaptive filters are used to optimize the audio after neural network enhancement to realize the enhancement of underwater target radiation signal. The simulation results show that when the SNR is -5 dB, the SNR of the proposed method is 17.51 dB. Compared with 1.23 dB of multiwindow spectral subtraction, 7.21 dB of convolutional denoising autoencoder and 4.12 dB of adaptive least mean square error filtering, the proposed method has a higher SNR gain.

Abstract:Mechanical fingertips can better perceive information about the motion state of an object, which is essential to achieve a stable grip similar to that of a human hand. In order to provide better perception of slip information of objects at the end of fingertips, this paper proposes a tactile slip signal recognition method based on discrete wavelet transform. Firstly, the information from the 25 contacts of the tactile sensor is fused using the correlation coefficient method; then the obtained information is wavelet transformed to observe the difference between the tangential force signal and the normal force signal in the frequency domain components. Finally, the discrete wavelet algorithm is used to differentiate between tangential and normal forces by setting appropriate wavelet coefficients and performing eigenvalue extraction through several trials. The experimental results show that the same wavelet coefficient threshold of 0.018 can still distinguish the pressure and slip signals well under the action of normal force of 3 N size, which is applied to three objects with different contours of spiky ball, smooth ball and bumpy ball to cause sliding respectively. The results of this paper can provide the basis and technical support for the robot to achieve stable grasping.

Abstract:In response to the low accuracy and reliability of temperature measurement standard for meteorological purposes, the accuracy and reliability of quantity value transmission are low. A verification scheme for the calibration result of the second-grade platinum resistance measurement standard device has been designed. This scheme takes the calibration result of a provincial second-grade platinum resistance measurement standard device as an example and verifies it through transfer comparison method and peer comparison method. The results show that the absolute deviation values of the corresponding calibration results between the device and the high-level measurement standard device are both less than 0.019 ℃, meeting the requirement of the transfer comparison method. The absolute deviation of the weighted arithmetic mean of the corresponding calibration results between this device and the four measurement standard devices participating in the comparison is less than 0.007 ℃, which also meets the requirements of the peer comparison method. Therefore, both methods have been validated. The results indicate that the second-grade platinum resistance measurement standard device can obtain accurate quantity transfer data, This scheme can ensure the accuracy and reliability of meteorological temperature measurement standard and quantity value transmission.

Abstract:In response to the poor corrosion resistance and short life cycle of the current steel cable, as well as the inability to meet the requirements of the current engineering applications of oversized spans, carbon fiber composites provide a new research direction for finding new materials to replace steel cable ties due to their excellent mechanical properties and stable chemical properties. This study uses a combination of experimental and numerical simulation to carry out the analysis and detection of eddy current fields in CFRP crack damaged structures. The results show that the coil is most sensitive to the detection of crack damage in CFRP cables when the excitation frequency is 1 250 kHz. The amplitude of the signal at the crack increases with the increase of the crack depth; it shows the trend of increasing and then decreasing with the increase of the crack width, and the amplitude reaches the maximum value when the crack width is 1.5 mm. Eddy current inspection technology can be effectively applied in the detection of damage defects in CFRP, providing a reference basis for the future use of electromagnetic eddy current technology to achieve in-service defect detection of CFRP cables.

Abstract:The low classification accuracy of motor imagery EEG signals (MI-EEG) of the left and right hands limits the development of related brain-computer interface technology. The motor imagery EEG signals of 16 healthy subjects were collected experimentally. A discrete wavelet transform (DWT) and convolutional autoencoder (CAE) based classification algorithm for motor imagery EEG signals were proposed. The EEG signal is converted into a time-frequency matrix using a discrete wavelet transform and input to a convolutional autoencoder network for the feature classification of EEG signals. The algorithm obtained better classification results when tested on both the experimental dataset and the public dataset. The accuracy of the three classification groups of rest-imagine left hand, rest-imagine right hand, and imagine left hand-imagine right hand was 97.36%, 97.27%, and 86.82% on the experimental dataset, and 99.30%,98.23%, and 92.67% on the public dataset. The discrete wavelet transform combined with the convolutional autoencoder network model outperforms other deep learning methods (CNN, LSTM, STFT-CNN) in classification applications of motor imagery EEG signals of left and right hand.