Abstract：With the development of economy and society, the Data Mining Technology runs through all areas widely. In which the forest decision of the classification algorithm has become a hot topic.Decision forest algorithm is statistics theory that combins the set of decision tree classification, can deal with noise and avoid over fitting surpassingly. This article mainly introduced the several classic methods of decision forest algorithm and their characteristics. Researching algorithms in domestic and overseas were analyzed and summarized systematically from the process of the construction of the decision forest. In the face of big data is described in detail application decision forest distributed computing process. By comparison,this article summarizes the applicable scope of various decision forest algorithm.

Abstract：Disseminating traffic information is one of important application, which relies on broadcasting protocols. Many new routing protocol are presented for VANETs(Vehicular Ad Hoc Networks). Firstly, there protocols are classified, and the existing representative routing protocols for VANETs are analyzed and summarized.Moreover, performance metric of the existing routing protocols is summarized, and a qualitative comparison with evaluation model, simulation platform and metric is given in this paper. Finally, possible directions of future research related to broadcasting protocol are proposed.

Abstract：A Ka-band integrated compact down-converter monolithic microwave integrated circuit chip is presented for synthetic aperture radiometer application. This work is designed using an image-reject mixer consisting of double balanced diode ring mixers and a four-stage RF low noise amplifier. The conversion gain of the chip is 17～20 dB, image rejection above 20 dBc, and noise figure of 2.6～3 dB for the radio frequency of 29～34 GHz.

Abstract：Based on nano-metal-oxide-semiconductor field-effect transistor(MOSFET) device structure, starting from the basic drift-diffusion equation, the subthreshold drain current model and gate current model were established. Especially, the frequency and bias dependence is explicitly included in the current models. Through comparative analysis showed subthreshold drain current model with the feasibility of scaling and gate current has followed and frequency-dependent. Meanwhile the simulation results of the model are compared with the extracted experiment results to verify the accuracy of the model.

Abstract：Accuracy in radix 4 decimation in frequency(DIF) algorithm is researched with different operation mechanisms. We analyze different data forms and roundoff rules of radix 4 operation unit with fixed-point, block floating point and floating point. A simulation model is established on MATLAB platform for operation mechanisms including fixed-point and block floating point. In this model, accuracy is measured by noise signal ratio. The result shows that, when the input is rand sequence , the noise signal ratio is effected by input length and scale and input width of algorithm. With the simulation result, we solved the spectral distortion with weak signal in practice. Consequently this scheme has better valve in practice.

Abstract：SM4 is the first commercial block cipher published in China, which has been widely used. The resource-constrained devices in IoT(Internet of Things) are limited by low processing power and small storage space，in order to solve this problem, designed and implemented a SM4 lightweight structure based on reutilizing resources and rescheduling procedures. Key expansion and the round function used the core module to generate round key and encrypt/decrypt by turns, and there is no need to store all the round keys required for the 32 round of iteration. Key store & detect module and real-time CK generate module, which can generating CK in positive or inverted sequence, can significantly improve the decoding efficiency. The results shows that with the advantages of occupying less resources and has higher FOM, the lightweight structure proposed in this paper is quiet suitable for the use of resource-constrained devices in IoT, and has broad application prospects.

Abstract：In this paper, the grey system theory is used to solve problems during the process of establishing welding point failure model, such as lack of data points, no obvious regularity, difficult to predict, etc. Three kinds of grey model of welding point failure are established, and a prediction about the later status of welding point health is made. The experiment results show that three kinds of grey models can predict welding point health. Predicted and the measured values are consistent. The prediction of even grey model is better than the other two models.

Abstract：With the development of automotive electronics, the sensor in vehicle is also increasing. The newest sensor based on PSI5 with flexible use and lowest possible implementation costs becomes popular in vehicle sensor communication. But, during the production development phase, how to verify it becomes a new topic. Simulator based on DSPIC33FJ128GP706 is designed, wich simulates the PSI5 communication of Freescale accelerometer sensor MMA5214LWR2, achieving the HIL(Hardware in loop simulator) of this sensor.

Abstract：The remote monitoring and serving system is designed for electric vehicle. The system realizes the uploading and storage of the vehicle real-time operating data, which based on data collection and remote wireless data transmission. The vehicle- mounted controller collects and sends data, the server stores, analyzes, deals with the data, while the client provides a vision for the user. As a result, the vehicle′s monitoring function is realized. Through the data analyzing, the server acquires the vehicle serving information，and sends the information to the vehicle-mounted controller. Thus, the remote serving function is realized. The system was tested on a A0-level urban car produced by Great Wall company, and performed well with practicability and stability.

Abstract：In order to develop vehicle collision system, a 24 GHz FMCW vehicle ranging radar system is designed. The system includes radar RF module，intermediate frequency signal processing module and digital baseband processing module. The radar RF sensor based on the hetero junction field effect transistor NE3514 is adopted as the radar RF module. The intermediate frequency signal processing module consists of high pass filter, variable gain amplifier and low pass filter. High pass filter is realized by using operational amplifier TLV2374，variable gain amplifier is designed by operational amplifier TLV2374 and digital potentiometer MCP42100，low pass filter is designed by operational amplifier AD8532. Digital baseband processing module completes data acquisition, signal feature extraction and system control. The vehicle ranging radar can achieve 90 m distance measurement.

Abstract：In recent years, wearable devices began to be concerned in the medical field gradually, and the emergency rescue task that the fire brigade assumes is becoming heavy increasingly, to ensure the safety of the fire brigade in the fire rescue operation has a very important practical significance. So a kind of ear-cliping wearable device based on fire safety is designed, the pulse wave signal is collected by the photoelectric conversion method, then the parameters of blood oxygen saturation and pulse rate are obtained by CC2640′s filter processing, then data is transmited through the wireless bluetooth,so the fire control command center can realize real-time monitoring of the physical condition of the firefighters. The device has the characteristics of small size, light weight , low power consumption and wearable, it can be used in fire, public security and other departments.

Abstract：Ultrasound bone density instrument is a good measure of osteoporosis for the human body. The instrument has the advantages of no radioactivity, low cost and so on. The design of bone density instrument based on ultrasonic sensor is discussed from hardware and software，including ultrasonic emission, ultrasonic reception, signal processing，the establishment of embedded platform, and the construction of the standard T value and Z value database. The feasibility of this design scheme is verified from the actual test results.

Abstract：Multipath effect leads to low precision of indoor positioning based on Received Signal Strength Indicator(RSSI). The physical layer Channel State Information(CSI) with high granularity can better describe the indoor multipath environment to improve the accuracy of indoor positioning based on fingerprint. Using clustering algorithm to extract CSI improves fingerprint discrimination in different locations. A simple and effective method is used to match the class in the positioning stage. The experiment result shows that the positioning accuracy is improved by 24% compared with the previous algorithm in the case of using a single beacon.

Abstract：Because of the low stability and the difficulty in controlling, Newton-Euler equation is used to establish a mathematical model for the four-rotor aircraft, the implementation of attitude algorithm is proposed too. The hardware control circuit is designed，which chooses the STM32 MCU as the core and is composed of accelerometer, gyroscope, magnetometers and so on. In order to get accurate data of angles, Kalman filter is used to filter the interfering signal from the data which is calculated from the accelerometer and from the gyroscope. To make the aircraft hover, fly forward, fly backward, turn left, turn right, etc., a three-way cascade PID controller is adjusted which can control three kinds of attitude angle, i.e., the roll angle, pitch angle and yaw angle. The flight tests indoor and outdoor show that the aircraft can fly smoothly.

Abstract：Drawings of ancient buildings are tenures of traditional Chinese culture, and need to be informatization and inpainting urgently. A novel model of inpainting the drawings of ancient buildings is proposed, which carry on the dictionary learning by K-svd algorithm, and filling missing pixels by known pixels in sparse domain，so as to implement the inpainting of drawings of ancient buildings and remove impulsive noise. The experimental results show that the algorithm can inpaint the image more effectively and decrease the RMSE, this method has better performance than other dictionary learning algorithm, and has good application potential and good application prospects.

Abstract：At present, USB 2.0 has been unable to meet fast reading of the high speed and large capacity data recorder. High speed read interface based on USB 3.0 is designed. The system takes a high speed and large capacity airborne recorder based on memory array, USB 3.0 is Slave FIFO interface mode,and FPGA of the recorder is as an external master controller. A FIFO is established in FPGA to realize cache and transmission of stored data. Finally, the data is uploaded to the host computer by the interfaces of USB 3.0. This paper introduces mainly the design of USB 3.0 reading interface hardware and its firmware program and FPGA control program，using GPIF Designer II and Quartus II software for simulation and verification. Test results show that the rate of USB 3.0 interface transmission speed is up to 120 MB/s，which meets the requirements of high speed read.

Abstract：In order to solve the problem of data centralized processing of train monitoring and control equipment(such as LKJ2000、TAX2、TCMS), a multi-channel serial data acquisition system of train monitoring and control equipment which run in the embedded Linux system is designed by analyzing the train vehicle equipment interface electrical characteristics and protocol data. Besides, the method of UDP congestion control is been adopted to improve network utilization. Simulation results show that the system can collect the data of TAX2, TCMS accurate and reliable in real time, and the accuracy of data acquisition is up to 100%. The system provides a powerful support for the realization of the data acquisition of train remote monitoring and fault diagnosis.

Abstract：Aiming at the problem that there were inherent time deviations among eight channel quantum state signals at the exit of the quantum key distribution equipment, quantum key distribution photon timing calibration system based on TDC-GPX was designed. The system completed photoelectric conversion, signal conditioning about quantum state signal. A high precision time interval measurement chip TDC-GPX was used to measure pulse signal has been conditioned. To meet eight channel signals indiscernibility in time, the system used FPGA for data processing and adjusted emission time of eight channel signals. The actual test results show that the measurement precision is less than 80 ps, calibration performance is good by applying to actual quantum key distribution devices, it can meet system calibration requirement.

Abstract：Target recognition system was designed with high luminous flux optical antenna, polarization module and interferometer. Through the polarization spectral imaging technology, the plurality of polarization direction of the spectral image is acquired simultaneously in the same field of view. Polarized spectral image of the object is recreated by image fusion algorithm. At different distances, the steel sheet and a certain type of gun for testing, target signal is acquired by the present system and the traditional non-polarized system. The results show that signal to noise ratio of target recognition system is high based on polarization spectra, and its detection distance is very far away. Compared with traditional methods, a composite image of the target system is clearly visible. It has a higher anti-interference, high field adaptability and stability.

Abstract：Real time spectral imaging applications are data and communication intensive algorithms. SoC architecture using NoC communication is the most appropriate solution as it brings flexibility, performances and immediate reuse of IPs cores. Tuning the SoC-based NoC architecture for a dedicated algorithm is a complex task as the design space is huge and experimentations take time. This paper proposes an emulation platform for the evaluation and exploration of performances of the architecture on FPGA. The emulation platform is based on the NoC and parameterized emulation blocks. Emulating the communication inside the FPGA gives precise timing performances compared to simulation. The evaluation is illustrated for a spectral imaging algorithm dedicated to art authentication on the architecture containing MIPS processors and the Hermes NoC.

Abstract：The properties of SET in DCVSL is studied by SPICE simulation. The propagation of the single side SET pulse in DCVSL is simulated. The results show that after two-stage DCVSL inverter, the single side SET pulse is basically eliminated. In asymmetric load and threshold voltage conditions, the pulse broadening effect of the double side SET pulse in DCVSL inverter chain is simulated. The results show that in the worst case, the SET pulse broadening effect in DCVSL is more apparent than in CMOS. And in the typical case, the double side SET pulse propagated in the inverter chains of DCVSL do not broaden.

Abstract：It is a practical design challenge on linear polarized to circular polarized transformation reflectarray, that spatial compensation phase and phase difference of quadrature polarization can′t be controlled independently. To solve this problem, a novel design method based on vectorial resultant of the electric field is introduced to absolutely independently control the two phase value, and it provides a new way of thinking about the design of linearly polarized to circularly polarized transformation reflectarray. Dual-stack three-parallel dipoles unit is presented to realize the independent control of the two phases. An X-band linear polarized to the right hand circular polarized transformation reflectarray using this unit is investigated, fabricated and measured. Experimental results demonstrate that the gain of 22.4 dB and cross-polarization level of -28 dB are obtained at 11.3 GHz, with 10% 3-dB axial-ratio and 1-dB gain bandwidth.

Abstract：In image processing, aiming at the given target detection and localization of the image, in order to improve the speed and reliability of target detection, this paper introduceds a new target detection algorithm based on image local steering kernel(LSK).Using the same size window of the query image traversal target image, the paper gets a series of sub-graphs the same size as the query image, then calculates the LSK of the query image and the target sub-graphs, the image feature information data matrices are obtained.In order to increase the computing speed, the method of local preserving projection is used to reduce the dimension of the data matrices. Further, the similarity judgment criterion is used to analyze the similarity between the query image and the target sub-graphs, the target area, which is similar to the query image, is the detection result. The experimental results show that the algorithm of target location is of high reliability and accuracy.

Abstract：In order to increase the throughput of Device-to-Device(D2D) pairs and overall system, full-duplex(FD) communication is adopted to D2D communication. However, the self-interference(SI) brought by FD is inevitable. This paper considers that the D2D pairs can share the same resource of one cellular user, meanwhile, co-channel interference will be caused between those D2D pairs. The paper proposes a graph theory-based resource allocation scheme so as to mitigate such interference under the precondition of the signal-to-noise-plus-interference-ratio(SINR) of the cellular users and D2D satisfaction, then uses graph coloring algorithm to coordinate resource allocation between D2D users and cellular users so as to maximize system throughput. Compared with either traditional half-duplex(HD) D2D communication or single D2D pair shares single cellular user’s resource mode, the simulating analyses show that the scheme can increase the communication quality of D2D pairs and enlarge the overall system capacity of cellular network efficiently.

Abstract：Physical-layer authentication based on the channel information can be used to complement and enhance traditional authentication, which uses the wealth of the wireless channel resources essentially. In this paper，a physical layer authentication method is proposed to suit to time-variant wireless channels. Each sampling values of separable paths is related within a block transmission time in time-variant channels, and the channel detection method based on the basis expansion model takes advantage of the varied mutually orthogonal basis functions and the constant basis coefficients to approximate the state of separable paths. Therefore, the authentication accuracy can be improved using the channel detection method based on the basis expansion model for physical-layer authentication algorithm under time-variant conditions. Simulation experiments in orthogonal frequency division multiplexing(OFDM) system proves the validity of the authentication method and that it makes 2～4 dB improvement compared to the physical-layer authentication using conventional Least Square(LS) channel sounding.

Abstract：In order to provide the users of long term evolution for railway(LTE-R) system with reliable wireless communication service, the channel state information need to be obtained by channel estimation. In high-speed mobility scenarios, the wireless channel is a time-frequency doubly-selective channel. In order to estimate the channel, a large number of pilots will be introduced. To tackle this problem, a pilot optimization strategy of channel estimation based on a distributed compressed sensing(DCS) method was proposed. First of all, according to the channel sparsity in the delay domain, the joint sparsity between basis function coefficients was excavated. And the estimating equation was decoupled. Then, a novel sparse pilot pattern based on the DCS theory which can remove inter-carrier interference was solved out. The simulation results show that the system performance of channel estimation strategy is significantly superior to the existing ones by the optimization of pilot pattern.

Abstract：Currently, elevator maintenance exists some problems such as non-standardized maintenance, non-timely maintenance, and so on which probably will cause elevator safety hidden dangers, the research and design of elevator maintenance system are carried out to solve the above problems. Android technology, RFID technology and intelligent terminal technology are integrated in this system. This application, developed on Eclipse IDE and JAVA language, is available in elevator maintenance, elevator maintenance records management and other functions. It is an important technical support to improve the quality and efficiency of elevator maintenance.

Abstract：Otsu algorithm，also called the method of maximum classes square error，is one of classical methods for image threshold segmentation. As generalization of 1D Otsu algorithm, 2D Otsu algorithm fully considers information of both the image gray and the neighborhood relationship among pixels, thus it is able to filter noise effectively. However, it is time consuming because of its huge amount of calculation. Concerning the problem, this article presents an improved fast 2D Otsu segmentation algorithm, which further cuts down the amount of computation by decomposing the original 2D Otsu algorithm into two 1D Otsu algorithm, constructing a new threshold recognition function through integrating inter-class variance with intra-class variance, and reducing dimension. Experiment results show that the improved method is superior to the other two methods in terms of segmentation efficiency and effect.

Abstract：With the development of computer technology, optimizing the image matching algorithm can enhance the accuracy of the image detection, this paper optimizes the design of image matching algorithm based on the curvature scale space of corner detection technology, and extracts the feature of image and normalize feature point based on curvature scale space image feature point. It′s helpful to improve the accuracy of image matching. The results demonstrate that it finally reaches the requirements of image matching by using the algorithm，verifying the effectiveness of the algorithm, improving the defect of over separation of feature points in image matching and enhancing the accuracy of image matching detection by 10.0%. The algorithm has a positive application value, it is worthy of promoting in practical application.

Abstract：K-anonymity is a common technique of information privacy protection. But K-anonymity must cause the information loss and how to improve the usability of anonymizing tables is the emphasis of K-anonymity. To solve the problem, this paper puts forward a kind of anonymous privacy protection algorithm based on generalized path —SPOLG algorithm. It introduces sampling with equal probabilities to find the generalized path whose information loss is little and raise the efficiency of data processing. Experimental results show that the algorithm could satisfy the anonymous requirement, at the same time, it is more efficient than Incognito algorithm to reduce the information loss of data released.

Abstract：Traditional Census+Hamming distance stereo matching algorithm often treated since the neighboring pixels equally, which usually resulted in a high mismatching rate because of lacking of sufficient matching information. So this paper proposed a Stereo matching algorithm using adaptive weight Hamming distance based on Census transform by introducing spatial distance between neighbor pixels and their center pixel, which enriched the information of matching images. In order to increase the noise immunity and texture information, the paper used distance between pixels of gradient image as the weight of costs aggregation. At the same time, it used sparse aggregation windows to reduce the complexity of the algorithm. At last， post-processing sub-pixel interpolation was introduced to reduce mismatching rate. The result of the experiments show that the proposed algorithm can not only improve the matching accuracy，but also reduce the complexity of the algorithm, and it is suitable for real time stereo matching.

Abstract：The frequency of the system can directly influence power quality and transmission efficiency on the magnetic coupling resonant wireless power transfer. The paper aims to present new aspects on transmission system, by calculating the harmonics generated by the high frequency inverter circuit respectively on the transmission system of the primary and the secondary current response to analyze the frequency on the efficiency and power quality for MCR-WPT. Finally, the hardware circuit is designed and the related research is carried out. The experimental results are in good agreement with the theoretical and simulation analysis, which proves the validity and accuracy of the proposed method. The results show that the system frequency is close to the resonant frequency the primary and the secondary current distortion rate is smaller and the transmission efficiency is relatively high.

Abstract：Aiming at the features and requirements when high-power supercapacitors were charged, high frequency switching power supply technology was applied to design a high-power smart charger with three-phase alternating current inputted. The main topological structure of main circuit consisted of a full-bridge converter which was driven by an IGBT driver IC and soft-switching technology was applied to control the leading bridge and lagging bridge. Not only did the charger have a CAN communication interface, but also a series of protection circuits were designed to avoid fault including under-voltage, over-current, overheat, short circuit, overvoltage and default phase, so a fault self-diagnosis system was provided. The charger saved charging process data and battery status data and displayed these data by a LCD, meanwhile users could query historical data if need, thus smart charging and automated management was realized by the charger. We tested charging stabilizing accuracy in different working mode and charging efficiency in different output power, the results showed that the charger can work reliably and has a high charging efficiency, so it is suitable for the charging of supercapacitors.

Abstract：In recent years, power line communications(PLCs) are widely used in the field of smart grid research, which can be used to provide communication capacity for power system in a convenient deployment. Power line communication access network(PLC-AN) is mainly used to control the singals between the substation and the user and provide Internet access to the family. After investigating the feasibility of using OR in PLC-AN, this paper designs a PLC-OR,which is based on PLC-AN, then uses the distributed method to solve a bit-meter per second maximization problem. Through simulation experiments, it is proved that the transmission time of PLC-OR is lower than that of the traditional sequential routing with the same level of reliability.

Abstract：Because of the high power LED driver required for current balancing, a multi-output LED driver with CLL resonant utilizing the charge balance of the block capacitors is proposed. The first stage circuit using BCM boost topology achieves PFC function, while the second stage using CLL resonant converter realizes the current balancing between LED strings. The principle of the CLL resonant converter is analyzed. This paper merges four outputs into one and adopts the fundamental harmonic approximation analysis method to obtain the voltage and current characteristic of the converter. The key parameters of the CLL resonant converter and Boost-type PFC is also presented. On the basis of above analysis, a prototype was built up. The experiment results prove that the circuit adopting the CLL resonant converter can efficiently drive the LED strings. Meanwhile, each LED has the good performance of current balancing.