Abstract:Aiming at the needs of bionic prosthetic motion recognition, a lower limb motion recognition method based on time-frequency generalized S transform and VL-MOBP neural network was proposed. First, time-frequency generalized S-transform was used to measure 4 kinds of surface electromyographic signals and knee flexion of the lower extremities of 22 male subjects aged between 20 and 40 years old, between 170 cm and 185 cm tall and weight between 50 kg and 75 kg. Using multi-resolution analysis of the frequency signal to obtain the time-frequency cumulative characteristic curve of the signal when the time and frequency resolution were good, then extracting the mean and standard deviation of the amplitude of the time-frequency cumulative characteristic curve as the feature vector, and using the VL-MOBP neural network to recognize the three movements of human lower limbs: Walking, standing, and sitting. The experimental results showed that the proposed lower limb movement recognition method can achieve good recognition results, with an average recognition accuracy of 96. 67%, which is about 56% higher than the wavelet transform and about 36% higher than the short-time Fourier transform. Effectiveness in motion recognition has been verified.

Abstract:Most of the existing channel model verification methods can only verify the first-order statistical characteristics of the fading model, i. e. the amplitude and phase characteristics of the signal envelope. Due to the complexity and diversity of the fading channel model, the existing first-order statistical characteristic verification methods cannot accurately classify the channel model. A comprehensive verification method of Rayleigh fading model is proposed. Firstly, the Rayleigh distribution is verified by the first-order statistical characteristics. Then the Rayleigh fading model is verified by the Doppler power spectrum distribution. The Doppler power spectrum density function of the complex sequence of the fading channel is extracted. The logarithm mean square error (LMSEE) with the theoretical Doppler power spectrum density is calculated, and the type of Doppler power spectrum distribution is determined by LMSEE so as to complete the verification of the given fading channel model. A large number of simulation experiments and physical verification are carried out. The input signal is passed through each fading channel model to get the output signal. The statistical distribution of the output signal is analyzed to verify the recognition performance of common Rayleigh fading models. The experimental results show that the recognition accuracy is more than 98%, which shows the effectiveness of this method.

Abstract:Since the traditional denoise method is difficult to fully remove the noise in the high intensity focused ultrasound ( HIFU) signal, it is proposed to use compressed sensing (CS) to denoise of HIFU echo signals. In the design of the observation matrix, the traditional Gaussian random observation matrix is improved to a sparse circular structured matrix, reduced time to construct observation matrix and reconstruct signal. Simulation experiments show that compared with the band-pass filter, wavelet denoise method and empirical mode decomposition (EMD) denoise method, the proposed method has higher reconstruction signal-to-noise ratio (RSNR), both reconstruction mean square error (RMSE) and maximum error (ME) are smaller. Using different methods to denoise HIFU echo signals obtained at different temperatures and extract the second harmonic excitation efficiency, it is found that the variance and fluctuation of the second harmonic excitation efficiency curve obtained by this method are smaller, which validate the denoise method superiority in actual signal.

Abstract:The Bayes Bootstrap method is widely used in the field of small sample prediction. However, due to the random value points that are not conducive to the prediction accuracy in the randomly generated self-service sample, the prediction deviation is large. In view of this deficiency, this paper proposes the Bayes Bootstrap & k-means method. In the case of having small sample failure data, use the Bayes Bootstrap method to generate self-service samples to expand the capacity of the original life data firstly, and then use the k-means method to perform data clustering analysis to remove outliers as much as possible and filter out more data points that meet the forecasting rules. Finally, the multi-chip module interconnection structure double stress accelerated life prediction is used as an example to verify the calculation method. Compared with the Bayes Bootstrap method, the prediction accuracy is improved by about 81. 44%, which has certain engineering significance.

Abstract:Aiming at the problem of clustering analysis and prediction of residential daily electricity load, a prediction framework based on the fine classification of residential power load patterns was proposed. In order to improve the quality of features used for cluster analysis, feature selection was first implemented based on BIC criteria. Then, the CFSFDP algorithm based on weighted pearson distance is used to realize the accurate identification of the shape of the residential electricity load curve. Next, the LSTM prediction network is improved by a fusion activation function method. Finally, the improved LSTM network is used to predict the finely classified residential power load patterns. The experimental results show that the forecast error index obtained by the method proposed is MAPE= 6. 6792%, which improves the quality of load forecasting and has a good effect in the forecast of residential electricity load.

Abstract:PPG signal acquisition system tends to be affected by motion artifacts, in order to improve the high signal-to-noise ratio of PPG signal and the accuracy of physiological parameters calculation, an algorithm for PPG signal artifact correction based on adaptive recursive filtering estimation of eigenvalue K is proposed. After removing environmental interference and power frequency interference from the collected PPG signal, the peak values of systolic wave and dicrotic wave of the signal without artifact interference are calculated, and the characteristic K value of each signal cycle is obtained. Adaptive recursive filtering algorithm is used to estimate the K value of the interfered signal, and the estimated result is taken as the true value of the K value of the interfered signal. Finally, according to the estimated K value, a new undisturbed signal is synthesized by Gauss function method to replace the part of the original signal disturbed by pseudo difference. Through blood pressure measurement experiment, the practicability of this algorithm in wearable system is proved.

Abstract:In order to solve the cycle slip problem for double difference observation sequence in Beidou satellite navigation system, wavelet transform combineda sliding window linear interpolation method is proposed. Using wavelet transform to scale the carrier phase detection sequence to extract high-frequency coefficients. The singular value of the high frequency coefficient can be obtainedat the cycle slip,then replacing singular values with sliding window bilinear interpolation, and coefficient reconstruction is performed. The cycle slip can be repaired better. The experiment uses this method to simultaneously repair the first and second layers high-frequency coefficients and to repair only the first-layer high-frequency coefficients. The results show that the number of repairs of the first and second layers simultaneously is less and higher repair efficiency.

Abstract:Communications hardware, signal and image processing need a large number of mathematical operation, and coordinate rotation digital computer (CORDIC) algorithm can quickly calculate the triangle, the hyperbolic, the natural logarithm and the square root function on the hardware, what’ s more, IEEE 754 standard is the most commonly used floating point numbers, so proposes a processing coprocessor of floating point arithmetic. The high radix adaptive CORDIC (HCORDIC) algorithm has the advantage of fast convergence speed. By designing the floating-point multiplier and floating-point adder for this algorithm, the architecture of floating-point coprocessor is designed to calculate various trigonometric functions and transcendental functions. This architecture can achieve faster convergence while reducing output delay and keeping low error. The design has been synthesized on the field programmable logic gate array(FPGA),the results show that compared to Xilinx CORDIC IP and other CORDIC architectures, it performs better in terms of output delay, maximum operating frequency, critical path and calculation accuracy, etc. It can be widely used in many calculation scenes and has a strong engineering value.

Abstract:In the VANETs, task offloading can effectively solve the problem of insufficient resource storage and computing resources of vehicles, but a stand-alone mobile edge computing (MEC) server is usually unable to meet the task offloading requirements in vehicleintensive areas. Aiming at the above shortcoming, a joint offloading method based on task urgency is proposed. When the vehicle offloads the task to the local MEC server, the local MEC server will send the task to the nearby MEC server to meet the task deadline according to the task urgency, deadline and the load of servers. The simulation results show that compared with the traditional scheme, the proposed scheme reduces the overall task failure rate of the whole system by 17%, optimizes the server load of the whole network and increases the reliability of the network.

Abstract:In the ICP-AES measurement system, the phenomenon of spectral superposition causes most of the spectral lines to be interfered by different degrees of overlap, which results in subsequent quantitative analysis errors. In this paper, the method of spectral overlap interference correction is explained and the evaluation function is established according to the mechanism of spectral superposition; By using the finite difference method to calculate the approximate second derivative and the distribution of the minimum value of the approximate second derivative, the minimum range of the characteristic wavelength of the neutron peak in the overlap spectral lines can be determined, and use the range as one of the initial conditions of the differential evolution algorithm. Then use the differential evolution algorithm to get the optimal solution of the evaluation function as the optimal eigenvector of the overlap spectral lines. The interference spectral lines and target spectral lines in the overlap spectral lines are analyzed by the optimal eigenvector. The effect of parameters NP and G on the performance of the differential evolution algorithm is tested by simulated overlap spectral lines data, and the feasibility of the spectral line overlap correction method proposed in this paper in the field of engineering is verified by measured overlap spectral lines data. The experimental results show that the differential evolution algorithm can be used to calculate the optimal eigenvector of the overlap spectral lines, and parse out the target spectral lines and the interference lines in the overlap spectral lines, thereby achieving the overlap spectral line interference correction, which lays the foundation for the subsequent quantitative analysis of element content.

Abstract:Aiming at the serious packet loss problem in the harsh marine environment of wireless sensor networks, a low complexity packet loss recovery with Reed-Solomon codes is designed and implemented. Specifically, the sending node analyzes the information data packet, it further generates and inserts a small number of redundant data packets using RS codes encoding, which can ensure the sequential transmission of the data packets, and generates redundant check data using fewer resource encodings. Based on the received information data packet and check data packet, the receiving node corrects and restores the lost data packet through the RS codes. The proposed method is tested on two complex transmission channels, terrestrial and marine, and the packet loss rate (PLR) is established as a metric for the packet loss recovery with RS codes. The test results show that the low complexity packet recovery method with RS codes can reduce the packet loss rate affected by channel fading, ensure data integrity under severe packet loss, and improve system reliability.

Abstract:Wireless local area network (WLAN) based positioning plays an important role in smart homes, indoor navigation and userdefined services. Proposed a seq2seq model based WLAN indoor positioning method. The method is based on the seq2seq neural network model, which is widely adopted in the natural language processing (NLP). The seq2seq model can learn the relationships of the time sequences in the signal domain and the coordinate domain. After carefully designed signal pre-processing, sample augmentation and reasonable loss function, the learned model can be adopted for positioning. According to the experimental results from our collected data, our method can improve positioning accuracy compared with some other neural network based methods, including the RFSM method, the denoising autoencoder (DAE) based method and the f-RNN method, by 23%, 11% and 20% respectively.

Abstract:Removing nighttime haze technique is a research hotspot in both digital image processing and computer vision, which is of great significance in the fields of aerospace, autonomous driving and traffic monitoring. Surveyed the domestic and foreign research on nighttime image defogging in recent years, summarized from the viewpoint of the traditional methods based on physical models and the method based on deep learning. After that, illustrated those algorithms in detail and then characterized their strengths and limitations, followed by comparison and analysis from subjective and objective evaluation respectively. Finally, future research directions were prospected and some suggestions were given.

Abstract:Aiming at the shortcomings of neural network recognition in static pattern recognition, a dynamic pattern recognition technology developed in recent years—hidden Markov model is used to analyze gearbox vibration signals. First, the statistical characteristics of the gearbox vibration signals in the time domain, frequency domain and time-frequency domain are extracted to form a 34-dimensional full feature vector. Trained a set of full feature-hidden Markov model libraries;then, through the principal component analysis technology, the full feature vector is reduced in dimension, and the first 7 principal components whose absorption information is more than 98% constitute the principal component feature vector. Another set of principal component-hidden Markov model library was trained. Two sets of independent model libraries are used for gearbox fault identification. The results show that the full feature-hidden Markov model library has 97. 9% accuracy for the identification of normal gears and gear broken tooth and 100% for gear pitting. The program takes 22. 328 s. The recognition accuracy of component-hidden Markov model library for gear pitting and gear tooth failure is 100%. The program takes 4. 879 s. Therefore, the dimensionality reduction processing of the vibration signal feature does not reduce accuracy of fault identification, but improves the accuracy of fault recognition, and greatly increases the speed of the program. This is of great significance for fault diagnosis of mechanical systems.

Abstract:The internal structure of Aerospace electronics device is complex, and there are many unknown factors in the process of the transmission of remnant material signal, leading to serious interference of data. In order to solve the problems of weak interference resistance in the process of the transmission of remnant material signal, easy to overlap interference signal and unable to identify accurately, the remnant detection method of aerospace electronic device based on multi-sensor data fusion is used. The same signal is used to extract by multi-sensor, then determine the weight of correlation function through correlation processing of pulse, and obtain the support matrix and weight moment array, then output the result of pulse group after data fusion. According to the result of the experiment, the multi-sensor data fusion method can effectively restrain the signal interference and significantly improve the detection accuracy of remnant objects.

Abstract:In order to extract robust facial features and improve the decision-level fusion of multi-regional features, a new expression recognition method based on neighbor smooth binary pattern ( NSBP) feature descriptor and weighted evidence fusion ( WEF) is proposed. First, a NSBP descriptor is proposed to encode the image by determining whether the gray values of the center pixels in the horizontal, vertical and diagonal directions are within the gray value range of two neighborhoods in each gradient. Then the initial basic probability assignments (BPA) of evidences are constructed based on the extracted NSBP texture features of the eyebrows, eyes, and mouth regions. Finally, aiming at the deficiency of Dempster-Shafer (D-S) evidence theory in conflict evidence fusion, a synthetic method of weighted evidence revision is proposed to realize the decision fusion of three regional evidences. Experimental results show that the recognition rate of this method on the Cohn-Kanade ( CK) database is 95. 25%, and the average recognition time is 765 ms, compared with other related methods, the effectiveness of this method is also verified.

Abstract:The disorderly charging of large-scale electric vehicles (EVs) will increase the operation risk of peak-up-peak to the power grid. In addition, as a mobile energy storage device, the disordered discharge of a large number of EVs will also have an important impact on the stability of the power grid. Therefore, it is necessary to guide the charge and discharge behavior of EVs in an orderly manner. First of all, the general loads of disordered charge and discharge of electric vehicles in a residential area are analyzed, and the daily load under different response is studied under the guidance of peak-valley time-of-use electricity price. On this basis, considering both the benefits of the divers and the power grid, the optimal charge and discharge model of the EVs are constructed, which takes the lowest charging cost of the EVs and the minimum peak-valley difference of the daily load in the community as the optimization objectives, and selecting the peak-valley time-sharing interval as the optimization variable. The optimal charge and discharge time intervals are found by Pareto-based optimization multi-objective genetic NSGA-Ⅱ and Pareto-based optimization particle swarm (PSO), respectively. The results by the different optimization algorithms are compared. Finally, the Monte Carlo algorithm is used to simulate and analyze the model. Simulation results show that users can reduce the charging cost to some extent through the discharge compensation, and NSGA-Ⅱ algorithm is better than PSO.

Abstract:In order to solve the LED recognition problem that the number formed by the factors such as illumination, background, and image distortion in natural scene, a recognition algorithm of LED-LeNet convolutional network is proposed. Firstly, the self collected LED light font data set was classified according to the number. Image data preprocessing includes image ROI operation, resolution adjustment to 32 × 32 and data enhancement. The network was reconstructed by convolution kernel, swish activation function and dropout regularization which referred to LeNet-5 network. The algorithm was verified by TST digital image database of traffic signal countdown collected in natural scene. The recognition accuracy of the algorithm can reach 99. 52%, and the recognition speed was 1 ms. The experimental results show that the algorithm has obvious advantages in recognizing LED light fonts after adjusting the network structure and convolution kernel parameters and changing the training strategy.

Abstract:Virtual hearing technology can reproduce the same hearing effect as real sound images, and may even optimize listening experiences. It involves psychoacoustics, signal measurement, computational simulation and other aspects. It has been widely applied in the fields of virtual reality, communication and navigation, hearing aids and psychological rehabilitation, and has attracted wide attention from researchers. An important application of this technology is the construction of a virtual auditory space, which requires the acquisition of personalized head-related transfer functions. Several methods for obtaining personalized head-related transfer functions have been proposed. To summarize and discuss these methods, we classify them into three categories: acoustic measurement, numerical calculation and approximate estimation based on human characteristics. Research progresses and important achievements within each category were then investigated. We analyzed advantages and disadvantages of various methods from the perspective of practical application, and then looked forward to new trends in technology development in the future.

Abstract:In view of the fact that the ash in the heated surface of the boiler will reduce the heat transfer efficiency and safety, uses the cleanliness factor as a healthy indicator to monitor the health of the heated surface of the boiler, and proposes a model that combines empirical mode decomposition (EMD) and long short-term memory (LSTM) to predict future boiler ash deposit. EMD can decompose a time series into a series of intrinsic mode functions which are stable in frequency domain, both LSTM has a memory function, it can learn to mine hidden long-term dependencies between time series, the combination of the two increases the accuracy of time series prediction. It is verified by simulation software that the model has satisfactory accuracy in the prediction of the ash condition of the heated surface of the boiler, and compared with two commonly used models, it was found that the prediction accuracy increased by 67. 7% and 59. 2% respectively and the feasibility and validity of the model are verified.

Abstract:In order to improve reconstruction accuracy to resolve temperature distribution reconstruction problems, a method for optimal sensor placement based on random forest algorithm is proposed. Denoting different measurement sites as different sample features, a series of different sensor placements and the reconstruction errors which are calculated by these placements constitute a sample dataset. A random forest model is setting up by the sample dataset, feature importance is also evaluated, then the optimal sensor placement is determined by feature importance. Simulation test and combustion test are set up to verify the feasibility and practicability of the proposed method. Testing data shows that comparing the original method, the proposed method can improve the reconstruction accuracy by at least 20%. Research results indicate that the proposed method has a good practical value, it also provides a new probe of using random forest algorithm to solve industrial problems.

Abstract:The struggle between modern electronic warfare radar jamming and anti-jamming is very fierce. The jammer produces multisystem and multi-type jamming signals, and the radar constantly improves its anti-jamming performance. Aiming at the comprehensive demand of radar jamming system for anti-interference performance test and evaluation of radar equipment, a radar reconnaissance and jamming system is designed and developed on the VST platform by using upper computer and FPGA technology, which can realize the comprehensive functions such as radar signal reconnaissance, multi-system jamming signal generation and radar standard calibration. The hardware scheme, reconnaissance and interference technology, FPGA function design and key FPGA module implementation are presented. The test results show that the system has high rf technical index, rich interference pattern, wide frequency band coverage, dexterity and versatility, and has certain application and promotion value.

Abstract:A large-scale closed-loop integrated Hall current sensor is designed based on the 0. 18 μm CMOS technology. The on-chip high-sensitivity Hall device is used to measure the magnetic field generated by the detected current and linearly output Hall signal. After linear amplification, offset cancellation, and proportional integral ( PI) regulation, the Hall voltage is compared with the triangular carrier to generate a PWM signal for driving the full-bridge power amplifier. The output current from the power amplifier is then sent to the secondary winding of the magnetic ring to form a closed-loop secondary side compensation, so that the Hall device in the magnetic ring air gap is in a zero magnetic flux state, thereby improving the accuracy of high current detection. Simulation results show that the designed integrated Hall current sensor achieves a high measurement range of 30 kA, the high accuracy level of 1, the low power consumption of 1. 08 W, and the low chip area occupation of 0. 2 mm 2 .

Abstract:Aiming at the problem of poor recognition accuracy caused by human trunk occlusion in current human motion recognition algorithms, an action recognition algorithm based on weighted three-view motion history image coupling time series segmentation was proposed. Firstly, in order to effectively describe the shape and spatial distribution of the action, motion history image ( MHI) is extracted from the video sequence. Subsequently, the Kinect camera was used to extract the depth image to obtain the outline of the human target's action foreground. In order to recognize the self-occlusion caused by body parts, the outline of action foreground was projected to three view angles (3V) planes to form 3V-MHI, which enhances the correct extraction of action. Using 3V-MHI, a MII for recording and observing trajectories was constructed, which overcomes the information limitation of single-view MHI. Then, according to the adjacent 3V-MHI, the energy and direction of motion are calculated by using temporal segmentation to detect the beginning and end of motion and output the result of motion. In addition, the gradient value of MHI was calculated as the weight corresponding to each plane, and the weighted 3V-MHI was obtained. Finally, the extracted histogram motion template was compared with the pre-established database to complete the action classification and recognition. Experiments show that the method can effectively solve the problem of selfocclusion and has high accuracy and robustness for motion recognition under complex background and illumination changes.

Abstract:Aiming at the problem that the heartbeat signal is easily interfered by respiratory harmonics and other noises and difficult to extract, a vital sign detection algorithm based on N peak capture is proposed. First, the radar signal received by the average cancellation method to filter out static clutter, and it uses the range gate selection algorithm to extract the body surface vibration signal; then it performs low-pass filtering and autocorrelation processing on the body surface vibration signal to remove random noise. On the basis of extracting the breathing frequency, its higher harmonics are suppressed, and M peak frequencies are captured in the heartbeat frequency band, then the peak frequency of the maximum number of occurrences of the heartbeat frequency band is calculated iteratively N times as the heartbeat frequency. Simulation results show that the algorithm has higher measurement accuracy and better anti-interference ability than the discrete Fourier transform (DFT) algorithm, and can be effectively used in the field of vital signs detection.

Abstract:When the smoke concentration of forest fire increases, the blurring degree of the corresponding image increases, and the total bounded variation gradual declines. Based on the characteristics of the variation, the difference between the boundaries can be effectively represented. Therefore, a detection method of forest fire smoke image is proposed based on total bounded variations. The objective function is extremum calculated with the idea of block stationary analysis, and the total bounded value is obtained. By comparing the total bounded variational value twice, the suspected smoke is extracted from the block result graph, and the fused clustering processing of feature data is used to obtain the final suspected smoke area. In order to get better smoke detection effect, it analyzes the motion characteristic of suspected smoke feature area, and the smoke area is judged by fusion, then the fire alarm is given. The algorithm shields the complex calculation of the static characteristics of the smoke. When the suspected characteristics of smoke are analyzed, the smoke can be accurately detected by only its motion characteristics, which avoids the errors caused by cumbersome calculation. The comparison and verification results show that the output of the algorithm is efficient and stable.