Abstract：Since the rain line stripes in the image have different shapes and sizes and are unevenly distributed, the rain density of the single neural network learning uneven distribution is weak, and the rain removal effect is not significant. This paper proposes a rain density sensing guide expansion network to remove rain from a single images. The network is divided into two parts. The first part is the rain density perception network classifying the images of different density rains(Heavy rain, Medium rain, Light rain). The second part is the expansion network guided by the joint rain density perception classification information learning different rain density characteristics details for detecting rain lines and removing rain. Experiments show the effectiveness of the method in the de-rain on synthetic and real data sets.

Abstract：The maturity of 4G network technology makes users′ business demand for operators higher and higher. How to maintain users and cater to users′ business needs through the study of user attributes, establish convenient and fast experience service means, and build maintenance and retention system is the most important thing for the future development of China′s telecommunication operators. This paper firstly analyzes the current situation of mobile users to maintain the development, and puts forward the user maintaining development attributes. Secondly, a data mining method based on data mining is used to analyze the data mining model, which is based on the user′s stability and user value. Finally, further prospects are put forward on how to carry out multi-channel precise push for stock maintenance.

Abstract：Aiming at the problem that the national network customer telephone voice recognition has poor recognition of core words in specific fields, this paper proposes a method based on HCLG domain weight enhancement and domain word correction, which can add domain words in real time and quickly, to dynamically optimize the language model and improve speech recognition. The model and algorithm are optimized in the various fields of the telephone voice consultation, maintenance, complaints, etc. of the State Grid Customer Service Center. The speech recognition results have been greatly improved.

Abstract：Automatic modulation recognition of the multi-system communication signals based on feature extraction and pattern recognition is an important research topic in the field of software radio. It′s one of the key technologies for a complex electromagnetic environment in the field of non-cooperative communications, such as spectrum management, spectrum detection. A new algorithm for communication signals automation modulation recognition based on deep learning is proposed in this paper. It utilizes the autoencoders for feature extraction to obtain feature set with high anti-interference ability, then classifies and identifies the selected features with BP neural network. The algorithm can realize the automatic identification for MQAM communication signal modulation. Simulation results demonstrate that the propsoed algorithm has a good performace in classification and recognition, meanwhile effectively improving the anti-interference ablility of the automatic identification of the digital modulation signal.

Abstract：Based on commercial 0.2 μm SOI RF process platform, the influence of stack, width, bias resistance and bias voltage on scatter parameter characteristics of test structure for SOI RF switch application is investigated, including series branch, shunt branch, single-pole-single-throw, and single-pole-double-throw. The impacts of different design on RF switch insertion loss and isolation are discussed in detail. The result can be reference as device optimization and circuit design.

Abstract：The principle of data power failure protection for non-volatile memory(NVM) of security chip is analyzed. Considering that the processing speed of software can not meet the performance requirements of security chip for data storage, this paper proposes a NVM data implemented in hardware for power failure protection solution. The scheme adopts a ping-pong structure and takes two Flash spaces of the same size in turn as the target area and the backup area. After each update is completed, the backup flag and the number of backups need to be updated. If the chip is powered off during the update, after comparing the backup codes and backup times of the two areas after power-on again, it can be determined which area of the data is valid. This scheme can not only ensure that the data of the security chip non-volatile memory(NVM) is not lost, but also can improve the performance of NVM data update.

Abstract：The unified rendering architecture GPU provides a rich graphics computing and storage resources and puts forward higher requirements for software optimization. To design and optimize performance effectively for the unified rendering architecture GPU，on the basis of research on the unified rendering architecture GPU architecture and working principle, this paper analyses the various factors influencing the graphics: graphic instruction generation, host interface data transfer, graphic instruction parsing, graphics processing pipeline and unified rendering array processing capability. In the process of development of GPU, using this method to design each part of the performance indicators, through the simulation shows that the model assessment graphics performance and the measured phase, the error is less than 7.5 percent.

Abstract：Aiming at the limitation of SOC′s pin number, and in security IC applications, on chip capless LDO and on chip decouple capacitor were preferred for suppling digital cores power. But On Chip Capless LDO′s finite bandwidth limited it responses sharp current pulses existed in digital core, huge on chip decouple capacitors of several nano farads are necessary for these situations. These on chip capacitors occupy significant chip area, so these capacitors must be mapped out in the whole chip floorplan and so did LDO designs. In conventional design flow, the on chip capacitor′s size was confirmed after P&R with power simulations at the late design procedure, which results in iterations of LDO design and whole chip floorplan distribution. A new accurate method was given based on deeply understanding digital circuit how to work in order to estimate the values of on chip capacitors at the early procedure of the flow, so the floorplan of the chip and LDO design could be started earlier, the iteration of these procedure was limited to one time. This method was succeeded applied in one SOC design, the post simulation and test results show that the estimated on chip capacitors are very accurate.

Abstract：According to the present situation that the high precision and high reliability demand of GPS/INSS integrated navigation system rose, a robust adaptive extended Kalman filter(RAEKF) method was proposed, considering the problem that the mathematical statistical characteristics of noise in motion process were not accurately determined due to the complexity of carrier motion and the filtering results were easy to be affected by the observed coarse difference. Based on inertial navigation, satellite navigation and error model, the mathematical model of Kalman filter under tight combination mode was determined. The EKF model was derived and the adaptive Kalman filter was constructed. The filter precision was improved by adjusting the use weights of the observed value. The results showed that the influence of gross errors on navigation results was effectively suppressed by RAEKF.

Abstract：This paper presents a new ultra-wideband(UWB) high-gain dual-band common-aperture array antenna. The antenna consists of four T-shaped dual-polarized antennas, a 6×6 waveguide slot array antenna, and a circular metal substrate. The dual-polarized antenna operates from 2.33 to 12 GHz and uses a large metal Vivaldi antenna array that sits above the metal plate. The waveguide slot antenna array operates in the X-band with 9.8～10.3 GHz and is located below the metal plate. The dual-polarization antenna array and the waveguide slot antenna array are located in the same plane and separated by a circular metal plate, which can improve the space utilization and can effectively reduce the mutual coupling between the antennas and solve the problem of poor isolation effect of common-aperture antenna. This article uses HFSS software simulating, the simulation results show that the common- aperture antenna array has the advantages of high polarization isolation, ultra-wideband, high gain, high aperture efficiency, and is suitable for modern communication systems and electronic countermeasures.

Abstract：Unified power flow controller(UPFC) is the most advance flexible ac transmission system(FACTS). Active/reactive power balance for UPFC is analyzed in this paper when it is working in power flow control mode, and an improved real time power coordinate control strategy is proposed in paper. In the parallel converter control loops, an improved real-time dynamic feed forward method is added to reduce over voltage caused by imbalance power. The results of hardware-in-loop simulation based RT-LAB platform show that this new coordinate control strategy，comparing with the conventional one, can decrease excessive voltage excursions of the send terminal ac voltage and DC-link voltage and improve the control performance during power flow changing or system occurring fault.

Abstract：For the lapped bi-orthogonal transform(LBT) in JPEG XR encoding algorithm, the traditional implement method is always using linear lift structure which has no timing constraint, and it could cause some problems such as metastable state, uncontrolled code flow. Therefore, a LBT transform controller based on FPGA′s timing structure has been designed. In this design, mixed state machines is used to make the LBT transform operators to be data dealing modules that are timing-controlling-structure. The data dealing modules communicates with front and back controlling modules through handshake signals, and processes data according to relevant instructions. Single RAM cycle structure is used in the design, so that it can save FPGA internal storage space. The channels between each controlling modules and RAM are switched via channel chooser. Each controlling module computes addresses of image data in real time and reads or writes the RAM crossways on the basis of the feedback instructions of data dealing modules. Experimental results show that the controller has realized FPGA timing constraint on LBT transform, and the images processed by it are almost same to the results on MATLAB. This design has achieved desired goals.

Abstract：This paper proposes a kind of design for satellite tracking platform which is based on software defined radio(SDR) technology and constructed with several properties such as open, modularization, code refactoring, etc. Compared with those platforms without code refactoring, the function of dynamic code refactoring in this platform behaves some characters of task switching, equipment reusability, configuration flexibility, high-speed. This platform can update and extend its function to meet different requirements for satellite tracking tasks.

Abstract：Two methods for measuring pulse wave using either a pressure or an optical sensor are studied and compared. A model for determining the optimal external force for restraining the pressure sensor is proposed. Pulse waves are measured on 20 volunteers. The restraint force on the pressure sensor is slowly varied and recorded using a piezoelectric film sensor for each person until it reaches a so-called optimal acquisition pressure under which the pulse wave has the greatest amplitude. The optimal pressure is then analyzed using polynomial regression with the body-mass index(BMI). In addition, the pulse waves are decomposed in both the time and frequency domains to identify and extract the features corresponding to physiological activities. Pulse waves measured with the optical sensor method suffer from feature point loss where certain points are submerged and indistinguishable. The pressure sensor method can reveal the feature points clearly, but the waveform exhibits variations if the external force applied isn′t a constant. With the application of optimal acquisition force set by our model, the latter method is able to detect pulse wave with a high signal-to-noise ratio. The dual domain analysis is shown to extract features that can be later used for continuous blood pressure monitoring.

Abstract：The terahertz wireless personal area network is a new type of central network without infrastructure, supporting data transmission rates of 10 Gb/s to 1 Tb/s, and has broad application prospects. The existing terahertz wireless personal area network(MAC) protocol and the efficient directional MAC protocol(ED-MAC) have the problem of large overhead and long time, so a fast and efficient terahertz wireless personal area network oriented MAC protocol FE-MAC(Fast and Efficient MAC protocol) is proposed in this paper. Through simulation verification, the proposed FE-MAC protocol is compared with the 802.15.3c-based terahertz MAC protocol and ED-MAC protocol, which can effectively reduce the data transmission delay between nodes and improve network throughput.

Abstract：5G Ultra-Dense Network(UDN) allows a large number of high-density low-power transmission points(TPs) to be deployed in heterogeneous networks(HetNets). It greatly improves the system performance whereas introduces inter-cell interference(ICI). To eliminate ICI, CoMP concept has been widely used in UDN. But it also brings new challenges, how to effectively allocate network resources is one of them. This paper proposes a CLA-CoMP algorithm to improve both the rate of the small cell edge users and the system sum rate. In order to deal with the power allocation problem in UDN, this paper adopts P2UPA strategy, which aims to allocate higher power to users with the high probability of association. Simulation results reveal that jointly use CLA-CoMP and P2UPA can significantly improve the system performance, compared with the case of average power allocation.

Abstract：A novel frequency reconfigurable microstrip patch antenna is proposed, which uses liquid metal alloy and highly extensible elastomer as a radiant patch and a dielectric substrate, respectively. The liquid metal used is eutectic gallium-indium alloy(EGaIn), consisting of 75% gallium and 25% indium. The elastomer used is polymethylsiloxane(PDMS), and the tensile quantity of PDMS under three ratios is measured by experiment. When the mass ratio is 10:1, PDMS has lower viscosity and better stretchability. The PDMS model of the antenna is made by soft lithography, and then the surface is processed so that it can be encapsulated better. Finally, the EGaIn is wrapped in the PDMS to form an antenna. In the tensile limit of PDMS, by axially stretching the microstrip antenna, the resonant frequency gradually decreases as the length increases, and the frequency can be reconfigurable in the range of 4~6 GHz.

Abstract：In order to overcome the problem that it is difficult to accurately define the downsampling operation for a generalized graph signal in graph filter banks, this paper focuses on the design algorithm of nonsubsampled graph filter banks. Firstly, the spline filters are taken as the analysis filter banks. Then, two different methods are proposed to construct the synthesis filter banks. In the first algorithm, the synthesis filter banks can be constructed with the perfect reconstruction conditions in vertex domain. By taking into account the frequency of the subband filters in the second algorithm, the synthesis filter banks are designed by solving a constrained optimization problem involving the spectrum characteristics of the filters. The design methods can lead to two channel nonsubsampled graph filter banks with perfect reconstruction. Finally, taking the two channel nonsubsampled graph filter banks as a basic building block, multichannel nonsubsampled graph filter banks are constructed，which can realize multiresolution analysis of graph signal through cascading. Simulation results show that the designed nonsubsampled graph filter banks have perfect reconstruction property. Furthermore, the designed multichannel nonsubsampled graph filter banks have better denoising performance than the existing graph filter banks.

Abstract：To solve the problem that when positioning the inner and outer edges of the iris, insufficient contour information may result in inaccurate positioning or failure，a positioning algorithm combining Hough transform and least squares method is proposed.Firstly morphological opening is used to reduce isolated small dots in the image,and adaptive threshold algorithm is used to convert images into binary images.Then the morphological closing operation is used to complement the small gap to smooth the edges of the image and Canny edge detection is used to extract image edge information.Finally, according to the difference between the inner edge and the outer edge, the inner and outer edges of the iris are positioned by Hough transform and least squares respectively.Verified by experimental comparison and analysis, this method can locate iris faster and more accurately than traditional algorithms.

Abstract：In 3/2 connection smart substation, the voltage and current sampling values of electronic transformers in different bays are usually transmitted to a digital power meter through a cascade of merging units or network. The synchronism of voltage and current sampling time influences accuracy of power metering. So it is a key problem to synchronize the sampling values of multi-bays. A new digital power calibration device is designed, which applies to 3/2 connection system. It receives simultaneously the merging units IEC61850-9-2 messages of the PT in the bus and the CT in different bays by multi-fiber ports. Also, it uses the information of the messages and Fourier transform to compensate power angle deviations caused by synchronism. According to test data, the new device can compensate power angle deviations and achieve accurate power metering in 3/2 connection system.

Abstract：Aiming at the sparse problem of object point cloud based on structure from motion method, a 3D reconstruction method using different matching data is proposed. The matching points are calculated by contrast context histogram(CCH) algorithm. The M-estimation sampling consensus(MSAC) algorithm is used to calculate the fundamental matrix, the translation and rotation matrix are decomposed from fundamental matrix. The image projection matrix is obtained combining the camera internal parameters. KLT algorithm is used to update the matching data, and the point cloud is generated by triangulation principle. This method makes use of the advantage of high accuracy of CCH algorithm to make the calculation results of the basic matrix converge. Using KLT algorithm to realize the matching by displacement instead of description vector, it makes up for the deficiency of matching data in low frequency region. The experimental results show that the proposed algorithm is effective and feasible, and the reconstructed point cloud has advantages in comparison with existing algorithms, it can be used for building 3D model of objects in the real scene.

Abstract：The process of landslides in mine dumps is a dynamic, large-delay, highly nonlinear characteristic problem. There are many factors affecting the landslide of mine dumps, and each characteristic index has mutual influence. But there is no strict division standard of landslide warning for dumping sites, this paper proposes a method of combining Adaptive Boosting(AdaBoost), improved Particle Swarm Optimization(PSO) and Extreme Learning Machine(ELM) for short-term prediction of mine dumps. Firstly, the particle swarm optimization algorithm is used to obtain the optimal input parameters of the ELM model. Then, the adaptive learning algorithm is used to form a plurality of extreme learning machine weak predictors into a new strong predictor and predict it. The data collected by the soil field is taken as an example. The results show that the improved particle swarm optimization algorithm, adaptive lifting algorithm and extreme learning machine model combination method have better prediction accuracy than the extreme learning machine model optimized by particle swarm optimization algorithm and separate one. The prediction accuracy of the extreme learning machine model is close to the true value, which provides a possibility to realize the landslide warning of mine dumps.

Abstract：In view of real-time demand for speed compensation on high speed moving platform, an implementation method of missile speed compensation is proposed. After the idea of range cell migration compensation applied to the missile speed compensation, the signal-to-noise ratio of coherent accumulation is improved, and hardware design is realized on FPGA. The simulation experiment shows that the missile speed compensation method with FPGA has advantages of high real-time, fast processing speed and high precision.

Abstract：Action recognition is an important area in computer vision. An action recognition system based on SoC FPGA and CNN model is presented. A pipelined LK optical flow module and a pedestrian detection module based on HOG and SVM are realized. The results of detection of dynamic pedestrian and static pedestrian are merged to recognize the area where the person is. The image and optical flow in the area are both input into the CNN model, to recognize what action it is. An ISA-NPU is proposed to realize the computation of the CNN model. The system is realized in Altera DE10-Nano Kit. With the co-operation of hardware and software, the system can recognize actions such as stand, walk, wave and squat. The system is accurate in recognition and flexible for expansion and modification of the CNN model.

Abstract：As the most important safety component of the construction elevators, the periodic detection of anti-falling safety devices is particularly important. The embedded detection system based on the Android platform of the SAJ30/40 gear progressive anti-falling safety device was designed, the angle sensor was used to measure the rotation angle of cone hub of the anti-falling safety device and the speed sensor was used to measure the real-time running speed of the elevator respectively, the data was processed by the upper computer analysis software, the braking distance and the operating speed of the anti-falling safety device were output. The test results of the method of the field test frame show that the detection system can measure the braking distance of the anti-falling safety device and other testing parameters accurately, and the measurement accuracy of the braking distance is ±1 millimeter.

Abstract：For the power factor correction converter circuit structure contains low-frequency rectifier bridge causing the problem of low efficiency, Boost PFC converter can only boost the pressure but can not step down，the proposed bridgeless Cuk PFC converter can convert high voltage to low voltage and increase converter power factor. Aiming at the problem that the converter′s ability to suppress the input voltage is poor and the dynamic response speed is slow, a single-cycle control method is proposed to improve the converter′s ability to suppress the input disturbance and the dynamic response speed. This paper introduces the derivation of the control equation for the single-cycle control of the bridgeless Cuk PFC converter and the design of the control circuit. Using MATLAB/Simulink to simulate the operation of the converter, the results show that the single-cycle control method of the bridgeless Cuk PFC converter can achieve good power factor correction, reduce the loss of the converter, and achieve good power supply interference suppression.

Abstract：A compound control strategy is proposed in order to solve the low control accuracy and poor anti-disturbance of deadbeat control. In this method, the repetitive control is embedded into the deadbeat control to improve the robustness of the control system and the variation of inductance parameters. According to the mathematical model of three-phase four-leg inverter, the deadbeat control algorithm applied to the current loop is deduced and modified, then the repetitive control algorithm is added. Finally, the strategy is simulated and verified by experiments to verify the effectiveness of the strategy.

Abstract：In the photovoltaic power generation system, the boost circuit for maximum power tracking has problems such as insufficient circuit boosting capability and large input ripple, an improved boost circuit to embed switching inductance into parallel interleaved boost converter is proposed. The circuit has four switching modes in the same switching cycle, where there are three different working states, and the output voltage gain and input current ripple at different duty cycles are discussed by using the average cycle modeling method. The result of MATLAB simulation shows that the improved boost circuit has higher boosting ability than the traditional boost circuit. In dynamic input condition, it has faster tracking speed, more stable output power and smaller input current ripple, which is suitable for the maximum power point tracking.

Abstract：Electrolytic capacitor as filter capacitor is widely used in switching power module, but it has higher failure rate than other components. In order to monitor the status of the electrolytic capacitor, the equivalent series resistance(ESR) is selected as monitoring target. This paper presents an online monitor method for buck converter output capacitor’s ESR and capacity value, by sampling output voltage at falling edges and rising edges of switching signal. The effectiveness of the proposed method is verified by the experimental system. The proposed method avoids measuring inductor current that simplifies the trigger circuit structure and keeps the topology of the switching power supply module, which facilitates a wide range of applications.

Abstract：In order to improve the grid-connected stability of two stage PV inverter based on boost circuit, the operation characteristic of PV inverter under unbalanced fault is studied. Combining with boost circuit regulating the DC bus voltage, a current control strategy based on negative sequence voltage feed forward is proposed to achieve LVRT. The conventional boost circuit voltage control mode and output current control strategy are analyzed. A DC bus voltage control outer loop based on PV array maximum power point voltage feed forward is introduced to change the boost circuit voltage control mode. By this way, the purpose of making the DC bus voltage stable and preventing over current of the inverter can be achieved. Then, a current control strategy based on negative sequence voltage feed forward is designed to ensure the output current three-phase symmetry under unbalanced fault. To provide a certain reactive power support to the grid, the current reference value is provided directly according to the voltage sage. In order to achieve the aforementioned current control method, a positive and negative sequence separation function module based on SOGI is designed. Finally, a 100 kW PV inverter PSCAD simulation model is built to verify the proposed control strategy. The result shows that the method not only can suppress the DC bus voltage rise rapidly and effectively, but also can make the inverter provide more dynamic reactive power support for the grid even when the single-phase voltage falls to 0.2 p.u.