Abstract:As an advanced function of the human brain, emotion has a great impact on people's mental health and personality characteristics.The classification of EEG emotion data sets can provide further theoretical and practical basis for real-time monitoring of normal and depressed patients' emotions in the future. The article uses the differential entropy features extracted from the public EEG emotion data set, and uses traditional moving average and linear dynamic system methods. Using the convolutional neural network in deep learning as the basic premise, a convolutional neural network's EEG signal emotion classification model is designed, which includes 4 convolutional layers, 4 maximum pooling layers, 2 fully connected layers, and 1 A Softmax layer, and batch normalization is used to make the parameter search problem easier and suppress the model over-fitting. The experimental results show that the average accuracy of the three emotion recognition of the SEED data set using this model reached 98.73%, the precision, recall and F1 score were 99.69%, 98.12% and 98.86%, respectively, and the area under the ROC curve reached 0.998. Compared with recent similar work, the convolutional neural network structure proposed in this paper has certain advantages for EEG signal emotion classification.

Abstract:In order to improve the stability of the Permanent Magnet Synchronous Motor(PMSM) vector control system and reduce the hardware cost, a bipolar switch Hall sensor with high stability and low cost is used to replace the high-resolution sensor in the control system to achieve precise control of the permanent magnet synchronous motor . The article first introduces the working principle of the Hall sensor, combined with the structure of the permanent magnet synchronous motor, introduces the application of the bipolar switch Hall sensor in the motor; then analyzes the installation phase deviation of the Hall sensor, and gives the calibration of the phase deviation Method: Reuse the construction of a Hall position observer to realize the detection and estimation of the motor rotor position and speed; finally, the feasibility and superiority of the proposed Hall installation phase deviation calibration method are verified through the construction of system simulation and motor test platform experiments . The method proposed in this paper can reduce the estimation error of the rotor position of the motor to 0.33% and the estimation error of the speed within 0.5%. At the same time, the estimation stability can also be maintained under load. Improved the robustness and reliability of the permanent magnet synchronous motor control system.

Abstract:In the practical application of indoor visible light positioning technology, the signal-to-noise ratio is reduced to some extent because of the effect of the ambient light. This is a challenge to the visible light positioning system based on the received signal strength. A visible three-dimensional positioning system based on sub-carrier modulation is studied. The interference of the ambient light on signals is decreased using sub-carrier frequency selection in the receiver. The identity information and received signal strength of LED is obtained by noncoherent demodulation. The target's location is computed based on an bare-bones particle swarm optimization algorithm. The three-dimensional positioning tests under different intensities of ambient light and different heights are carried out in a space of 1 m×1 m×1.89 m. The results show that the average three-dimensional positioning error of all test points is around 3.00cm, and the positioning errors are 2.83, 2.98 and 2.83 cm respectively when the height is 0.21 m and the illuminations of ambient light are 0, 10 and 30 lx. Experiment shows that the visible light positioning system based on subcarrier modulation can effectively reduce the positioning error caused by ambient light.

Abstract:To improve the decoding performance, based on the (8176, 7154) LDPC code applied to near-earth space in CCSDS standard, and according to the normalized minimum sum (NMS) decoding algorithm, design and implement the LDPC decoder. The design of the decoder mainly optimized the quantization data of check nodes, the scale factor changes with the number of iterations, and the scale factor value is based on the multiple of 2 and used the right shift addition to replace the multiplication of check node data and scale factor, which simplifies the hardware implementation. In addition, add a decoding verification module to test whether the codeword is successfully decoded after iterative calculation of the check node and the variable node, and the data is sent out after successful decoding or reaching the set maximum number of iterations. The LDPC decoder is designed and implemented based on FPGA. In the hardware design, used parallel decoding circuits to make rational use of hardware resources. When the signal-to-noise ratio is 1.8 and the maximum number of iterations is 15, through simulation and board-level verification, and comparing the decoding results when the scale factor value is 0.5, 0.75 and the scale factor is variable, it is proved that the variable scale factor NMS decoding algorithm can realize the decoding function and has good decoding performance.

Abstract:At present, the identification of malicious encrypted traffic based on machine learning mainly uses supervised learning and relies on a large number of labeled samples. However, in the real environment, malicious traffic is not only scarce but also depends on expert experience, and the labeling cost is high. Active learning selects difficult samples through iterative for training, which reduces the amount of training samples to a certain extent, but the current hardsample selection strategy based on committee votes has a coarser granularity, and the quality of the selected samples is not good. In response to this problem, a CBU (Committee-based Uncertainty, CBU) is proposed to improve the Query by Committee (QBC) method for identifying malicious encrypted traffic. Labeling sample similarity analysis, effectively measuring sample uncertainty, and selecting high-quality hardsamples to reduce sample labeling and training volume. The experiment uses the industry standard data set CTU and real malicious data sets for testing. The results show that compared with the traditional committee voting strategy, the amount of CBU sample labeling is doubled, and the recognition accuracy rate of only 15% of the data amount is 96%, which effectively reduces the sample labeling. And training volume, and it has strong practicability.

Abstract:In order to solve the problems of boundary mutation, poor stability and difficult determination of modal parameters in mode conversion of shearer mining height control system based on human-simulated intelligent control, a human-simulated intelligent control model based on fuzzy logic to improve mode switching and particle swarm optimization to optimize parameters is proposed. The model extends the characteristic mode of the human-simulated intelligent control error phase plane to a fuzzy set. The optimal control mode is selected in real time by fuzzy logic reasoning following the error and error change, and the control mode parameters are adjusted in real time by particle swarm optimization following the error and error change. The step response simulation is used to simulate the control performance of the fault at the coal-rock interface. The simulation results show that the human-simulated intelligent control model proposed in this paper improves the stability time by 4.71 s, the rise time by 0.345 s, the peak time by 0.671 s, and the overshoot by 5.458 % compared with the original human-simulated intelligent control algorithm. The stability, rapidity, robustness and parameter optimization of the model proposed in this paper are better than those of other control models in system modal transformation, and have superior performance.

Abstract:In view of the different characteristics of the frequency response of different audio playback devices, the Equalizer algorithm for audio frequency equalization adjustment and noise elimination is designed and implemented on the DSP. Based on the principle of time-domain filtering in digital signal processing, the IIR low-pass and high-pass filters that can pass through different frequency bands is designed. Different frequency bands are set with different gains. The output signal after the input audio passes through the IIR filter is superimposed to realize the audio regulating effect of energy changesat different frequencies; at the same time, the separate and controllable noise suppressor is designed, which is composed of a 50-order FIR band-stop filter for noise attenuation at a specified frequency. The test shows that the 16Bit digital audio signalup to 16KHz can be processed, and the filtering of noise in a specific frequency band is achieved.

Abstract:It is very important for single frequency network to realize high precision clock synchronization in Iot module connected to power system by power line carrier.However, the traditional clock synchronization method is not suitable for clock synchronization under power line carrier because of its low synchronization accuracy. In order to solve this problem, the hardware and software are studied. On the hardware, the high precision clock adjustment of physical layer is realized by using fractional phase-locked loop with high resolution output frequency.Software on the application of fuzzy control theory, the clock synchronization algorithm based on interval, to collect and analyze each receives the timestamp, reasonable adjust clock variables, study score type phase-locked loop to equipment clock frequency adjustment rules, to calculate the optimal clock adjustment parameters, realized the fast STA clock and CCO clock synchronization,keep the clock stable and keep the clock error within, significantly improve the clock accuracy, the accuracy and stability requirements of the Iot clock are fully met.At the same time, such high-precision synchronous clock provides technical guarantee for the application of Iot module in single-frequency network.

Abstract:The high-precision requirements of space gravitational wave detection and the limitation of laser line width and adjustment bandwidth put forward higher requirements for the design of the controller for realizing the optical phase lock of the response intersatellite laser interferometer. However, how to improve the system's suppression of laser phase noise when the control bandwidth is limited is a problem to be solved at present. In this paper, an internal model extended output feedback LQR controller is designed to improve the suppression of the phase noise of the laser under the premise of ensuring the stability of the phase-locked system and no cycle slip phenomenon. The simulation experiment results show that compared with the existing PI controller, under the control bandwidth of 20KHz, the controller designed in this paper can improve the laser phase noise suppression effect by 3 times. This controller reduces the control system's constraint on the maximum line width of the laser.

Abstract:This paper aims at the characteristics of large amount of test tasks and long test cycle of the large-scale phased array radar antenna, A high efficient multi-task plane near-field measurement system is introduced, based on traditional near-field measurement system, through adding a multi-functional signal simulator、single pole multiple throw RF switch and a dual polarized antenna with high isolation, the simultaneous test of the multi polarization and multi task comes true. Compared with conventional multitasking test system, the results show that the measurement time can be reduced by 200% or more, the measuring efficiency has been greatly improved, and the test accuracy can also meet the requirements.

Abstract:The routing protocol is of great significance to the quality of service (QoS) of the unmanned aerial vehicle self-organizing network (FANET). The high-speed movement of nodes in FANET will bring about drastic changes in the network topology. This feature will aggravate the link interruption, resulting in a significant reduction in network QoS. This paper proposes a low routing overhead protocol based on link quality (LQLR_OLSR). It optimizes the multipoint relay (MPR) set in FANET to reduce redundant MPR nodes; Then modified the expected transmission count (ETX) based on it as a routing metric, Integrated four characteristics of forward transmission success rate, reverse transmission success rate, link packet size and link bandwidth to realize routing multipath adaptation. The OPNET simulation results show that LQLR_OLSR in FANET environment is significantly better than Global_OP_OLSR and OLSR in terms of average throughput, packet transmission success rate, routing overhead and average end-to-end delay performance.

Abstract:Accuracy and real-time measurement of slurry density in recycling box in wet desulphurization pulping system are important for the economic and stable operation of desulphurization process, a prediction model of slurry density in recycling box based on improved sparrow search algorithm optimization (ISSA) least squares support vector machine (LSSVM) is presented. Secondary variables that are highly correlated with the slurry density are selected and preprocessed through mechanism analysis, and use PCA algorithm to reduce dimension. Chaotic mapping and adaptive weights are added to the standard sparrow algorithm (SSA), which improves the uniformity of population distribution and searching ability of the algorithm. It is used to optimize the key parameters of LSSVM and to achieve accurate prediction of serum density. The simulation results of actual data have shown that the average absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE) of ISSA-LSSVM measurement model are reduced by 44.5%, 43.8%, 43.9% compared with SSA-LSSVM, and the prediction accuracy is significantly better than that of the pre-improvement prediction model, which has some engineering application value.

Abstract:Aiming at the problems of the current TDOA (Time Difference of Arrival) unit positioning system of ultra-wideband, such as multi-solution of position and low positioning accuracy. In this paper, a cooperative positioning method of adjacent units and a cooperative positioning equation solving algorithm based on improved salphus swarm algorithm are proposed. In this method, by summarizing all the measured values of the same positioning request signal received in adjacent units, the coordinate solution is no longer limited to one clock synchronization unit, which also leads to an increase in the uncertainty of the number of measured values and the layout of base stations involved in the position calculation. Taylor algorithm cannot meet the solution requirements, so the improved salphus swarm algorithm is used instead of Taylor algorithm to calculate the cooperative positioning position, and finally the position coordinates of the point to be measured are obtained. The experimental results of the cooperative positioning of adjacent units of six base stations show that the R95 value of each measuring point is between 12 cm and 15 cm, and the maximum residual error is between 16 cm and 23 cm. The method proposed in this paper does not need multi-parameter judgment to choose or reject position information, which effectively solves the problem of multi-solution of position and improves the positioning accuracy.

Abstract:Considering the limitations that the current positioning methods havepower consumption, accuracy, etc., a NB-IoT positioning algorithm is proposed. Cloud computing is used to avoid the contradiction between the complexity of terminal equipment and power consumption; Base on the RSSI ranging model, the characteristics of wide coverage of the NB-IoT are used, and with the support of the reference terminal, the optimal strategy is used to select the relevant environmental parameters to make the ranging results more accurate. While using the least squares algorithm to obtain the target position, the constraint circle constructed by the relevant information of the reference terminal is used to determine and correct the final estimated position. Simulation and experimental results show that this method is feasible to improve the positioning accuracy; compared with general positioning algorithms based on signal strength, it has significant accuracy, and the confidence probability within 20 m error distance reaches 59.6%.

Abstract:This article focuses on the shortcomings of the bottom-up human pose estimation network OpenPose model with large parameters, and improves the feature extraction network and prediction network of the OpenPose model to achieve the goal of lightweight model.This article uses the ResNet18 network which has fewer parameters and higher accuracy to replace the VGG19 network in the original model.In order to reduce the amount of parameters of the network structure, we replace part of the convolution kernel in the prediction network with the deep separable convolution without losing too much recognition accuracy. Then, the human body actions are classified through the artificial neural network, and the linear module is added to the traditional nonlinear network to improve the memory and generalization ability of the network. The results show that the FPS of the lightweight OpenPose model has increased by 9% to 16% compared to the original. After 3000 iterations of the network training, the recognition accuracy of standing, sitting, walking, sitting and standing up reach 0.877, 0.835, 0.793, 0.815 and 0.808, respectively. Finally, the recognition network is applied to a real scene. According to the results, it is shown that the method in this paper runs normally in embedded devices and performs well.

Abstract:In order to solve the problems of slow smooth control response, poor adaptive ability, and weak anti-interference ability of the UAV flight control system,this article analyzes the principle of UAV flight control system and establishes a drone flight control model, under the condition of cascade fuzzy PID control, using the iterative optimization ability of PSO algorithm, the quantization factor, scale factor and initial PID parameters in fuzzy control are determined. PID parameters are adjusted online through fuzzy control, so that the parameters in the smooth control are kept optimized at all times.Therefore,a cascade fuzzy PID flight control system based on particle swarm optimization is designed. The experimental results show that: When the values of the quantization factors e and ec are 3 and 0.75, and the values of the scale factors k1, k2, and k3 are 0.5, 2, and 0.5, respectively, the system stability is optimal.compared with cascade PID control and cascade fuzzy PID control, the control system optimized by PSO has better control accuracy and stability, and can better improve the performance of the system and meet the flight requirements of fast and efficient leveling.

Abstract:Polarized Synthetic Aperture Radar can work in multiple polarization modes. Using multiple polarization echo data to achieve ground feature classification is an important application of polarization data processing. At present, there are still some problems in the application of convolutional neural network in the field of polarization feature classification. Including the information redundancy and dimension disaster caused by multi-dimensional polarization decomposition feature information To mitigate these problems, this paper proposes a fully convolutional network model based on feature fusion. Firstly, By designing a full convolutional network structure with two encoding layer branches, the deep features are extracted for the polarization decomposition feature and polarization scattering feature respectively to realize the separation of multi-dimensional feature information. Then, the attention feature fusion mechanism is adopted to realize the feature fusion of two branches, and the learning ability of the network is redistributed by sharing the attention weight of the connection layer learning channel. In addition, an improved Atrous Space Pyramid Pooling is introduced to improve the multi-scale prediction ability of the model. The experimental results show that the overall accuracy of polarization data sets in two different regions is 96.43% and 99.60% respectively, and the prediction time is 17.3s and 10.1s. The classification accuracy is improved without greatly increasing the prediction time, and the effectiveness of the algorithm is verified.

Abstract:A machine vision based drogue detection and location algorithm was proposed for getting exactly drogue detection and location results in docking process of probe and drogue autonomous aerial refueling (AAR). Firstly, reliable drogue feature and classifier models were generated through Adaboost training based on numerous positive and negative samples; then, using the drogue model, the drogue can be detected on the drogue image sequences according to machine learning based detection and tracking technology; finally, used the image domain position of drogue, the three-dimensional (3-D) position of drogue relative to refueling probe can be calculated through a mapping model from 2-D image to 3-D space. Simulation results showed that the drogue detection rate was over 95%, time elapse was less than 4ms/frame, and the position measurement deviation was less than 7%, which can meet the requirements of AAR.

Abstract:When the traditional image processing method detects the wear area on the tread of the wheel, due to the influence of the shadow and stains on the surface of the wheel, it is easy to cause misidentification. Ways to identify areas of wear. First, use a CCD camera to collect the low-speed wheel tread profile map, and then calibrate the worn area in the profile map to make labels, and use FCN-32S, FCN-16S, FCN-8S models for training. The experimental results show that the FCN-32S, FCN-16S, and FCN-8S models can effectively detect areas with large wear, and the FCN-8S model is significantly better than FCN-32S and FCN-16S for detecting point wear areas. , And there is no misrecognition phenomenon for the three models of the area with stain interference set in the experiment. Finally, the detection effect of FCN-32S, FCN-16S, and FCN-8S is evaluated by the MIoU value, and the number of model training is changed, the MIoU value will eventually stay near 0.7, and the detection effect is good.

Abstract:Visible infrared person re-identification is a cross-modal retrieval problem. Being able to accurately match pedestrians remains challenging due to the large modal differences between visible and infrared images. Recent research has shown that using pooling to describe local features of body parts as well as global features of the human image itself can give a robust feature representation even when body parts are missing, but simple global average pooling is difficult to obtain detailed features of pedestrians. To address this problem, this paper proposes a new global multi-granularity pooling approach that uses a combination of global average pooling(GAP) and global maximum pooling(GMP) to extract more background and texture information of person. In addition, the traditional triplet loss does not work well for cross-modal person re-identification. We design a new cross-modal triplet loss to optimise intra-class and inter-class distances and supervise the network to learn differentiated feature representations. In this paper, we experimentally demonstrate the effectiveness of the proposed method and achieves 88.01% Rank-1, 79.26% mAP, and 60.24% Rank-1, 57.50% mAP on the RegDB and SYSU-MM01 datasets, respectively.

Abstract:There are many types of optical lenses, and there are many objects to be tested on each type of lens. Traditional image processing methods have low measurement accuracy and poor adaptability to multiple types of workpieces. In response to the above problems, an optical lens measurement method based on machine vision is now proposed. Firstly, the corner points of the workpiece are located based on the improved Harris corner point detection. Secondly, the geometric primitives between the corner points are fitted by the least square method. Finally, the index information of the template size of each type of target is extracted, and the geometric basis is measured. The size between yuan. The experimental results show that this method can not only target a variety of workpieces, but also can perform high-precision measurement on workpieces with burr edges. By comparing the deviation of the measurement in this paper with that of the Keyence three-dimensional measuring instrument, the paper is obtained by analyzing the deviation and calculating the mean square error. The accuracy of the method is ±0.02mm, and the mean square error of the repeated measurement deviation is within 0.01mm.

Abstract:In order to solve the problem that the number of high-quality frequency standard for atomic clock output is too few, and in view of the demand for high-quality frequency standards for space electronic systems, this paper designed a novel space power divider based on the Allied Standard Avionics Architecture Council (ASAAC) standard. The design idea of power division first and then isolation is adopted, it not only has the distribution function of frequency reference signals, which can output 12 frequency standards with high isolation. But also the fault detection and communication functions can be realized by adding the control module and design of the built-in test system. In addition, the module adopts the main and backup design, which improves the reliability of the product. The switch of main and backup is completed by the output compare (OC) instruction. The LRM connector is used to realize the rapid plug-and-pull replacement of the module and enhance the maintainability of the module. Finally, Through theoretical analysis and measurement results, the isolation is better than 72 dB, the output power is greater than 6 dBm, the harmonic is better than -62 dBc, and the clutter is better than -90 dBc (within 200 MHz), which verifies the feasibility and effectiveness of the product.

Abstract:In view of the short time and poor stability of pulsed electromagnetic field, measuring electric field and magnetic field respectively will reduce the accuracy. In this paper, a scheme for measuring pulsed magnetic field and electric field at the same time is designed. Firstly, the relationship between electric field and magnetic field is deduced by using theoretical knowledge, then the half wave symmetrical antenna is modeled by cst2020, and the load effect and size effect of the antenna are simulated to determine the shape of the antenna. Finally, the signal is processed by charge sensitive amplification circuit. The simulation results show that the length of half wave symmetrical antenna is 15mm, the load capacitance is 1pf and the load resistance is 1m Ω, which is the best length. The charge sensitive amplification circuit processes the signal, the pulse current signal range is 0.1ua-1ma, the measurement range of electromagnetic field is linear, and the error is about 0.77%. The simulation results show that the shape parameters of half wave symmetrical dipole antenna and the waveform of pulsed electromagnetic field measured by charge sensitive amplification circuit can meet the requirements.

Abstract:For adults’ facial appearances changing slowly, the age estimation of adults in adjacent age groups is still a challenge. Aiming at this problem, this paper introduced the adversarial training method into the age estimation and proposed an age estimation method based on Adversarial Dropout Regularization(ADR). The age feature learner and the discriminator are trained via the adversarial training method, then the ability of age feature learning(especially the adjacent age groups features) gets improved. Experimental results on three classic datasets (UTKFace, MORPH and Adience) show that the proposed model improves the accuracy of UTKFace from 42.8% to 81.6%, and improves the accuracy of MORPH from 39.8% to 69.8%. Moreover, the accuracy of Adience is 63.3%. Being compared with other 4 classic models, the model in this paper using the neural networks of 5 layers achieves better results than other deep neural networks, and outperforms other methods with averagely 17.5% higher accuracy for the young and middle-aged(15-53 years old), which shows that our model improves the performance significantly on age estimation task, and has the practical value.

Abstract:In view of the problems of low accuracy, poor repeatability and large testing limitations in traditional motor-driven mechanical gear testing methods for smart wheel speed sensors, a smart wheel speed sensor test system based on LabVIEW is designed. The system adopts the static electromagnetic excitation sensor test method of non-mechanical components and innovatively designs a rotating magnetic field generator module, which can simulate the rotation speed and direction of any gear rotation, and realize the test of the wheel speed sensor at extremely low and extremely high speeds. Experimental research shows that the designed test system can test the functional parameters such as the high, medium and low current value, frequency and duty cycle of the output signal of the smart wheel speed sensor and analyze the sensor protocol information. The output frequency range of the tested sensor can reach 1~3 KHz, which improves the accuracy of the test system. The hardware platform of the system is built by modularization and has strong generalizability; the software adopts LabVIEW, simple graphical operation and display interface, which is convenient for testers to monitor and analyze the sensor output characteristics and adapt to the production requirements of modern wheel speed sensors.

Abstract:To solve the problem of low diagnosis rate of multi-type faults in industrial processes, a method of boundary discriminant projection (MDP) and support vector machine (SVM) fusion (MDP-SVM) was proposed. Boundary discriminant projection is often used in the field of face recognition, which can reduce the dimensionality of multiple types of data to obtain clear boundaries of different categories. Compared with principal component analysis (PCA) and local linear embedding (LLE), the local and global structures of samples are considered and the problem of small samples is avoided. The classification of dimensionality reduction data is judged by SVM classifier, and the optimal SVM classifier is obtained by particle swarm optimization (PSO) algorithm to achieve fault diagnosis. The simulation results show that compared with the traditional method, the fault identification accuracy of the proposed method can reach 95.379%, and multiple faults can be identified simultaneously.

Abstract:Aiming at the problem that the weak fault features of rolling bearings are difficult to extract, a bearing fault feature extraction method based on the combination of parameter adaptive optimization variable modal decomposition (VMD) and multi-point optimal minimum entropy deconvolution (MOMEDA) is proposed. Firstly, the VMD decomposition is performed on the rolling bearing time domain vibration signal, and then the best mode component (BIMF) is selected based on the principle of maximizing the index of impulse harmonic noise ratio (AIHN) of autocorrelation function and MOMEDA filtering is performed on it, and the fault characteristic frequency is obtained after envelope deconvolution, and finally the fault characteristic frequency can be clearly observed by applying the proposed method body to the numerical simulation signal 131.1Hz, which can be applied to the actual bearing fault signal to effectively identify the bearing fault characteristic frequency of 294.5Hz, which is closer to the theoretical fault characteristic frequency 294Hz compared with 311Hz extracted by the original envelope spectrum and 320Hz extracted by MCKD.

Abstract:The biggest advantage of finite set model predictive control is that the objective function increases the flexibility of constraints, but the weighting coefficient is difficult to determine and the instability caused by the coupling effect between multiple constraints greatly limits its application. To solve this problem, a multi-constraint model predictive control method based on Lyaponov control is proposed. This method first realizes the control of the output current of the main constraint through Lyapunov control, and then sets the weighting coefficient freely according to the weight of the constraint term of the switching times, and then realizes the multi-constraint cooperative control by minimizing the objective function. The simulation results show that the proposed method realizes the multi-constraint model predictive cooperative control of output current and switching times, and has good robustness to the weighting coefficient. When the deviation of the weighting coefficient is as high as 10 times, the output current THD is only 5.63%.