To solve the low recognition rate and complex algorithm of human body posture in WBAN(Wireless Body Area Network), a hierarchical theory is employed to design a human body posture hierarchical recognition algorithm. Considering the comfort level of the user, the nine axis acceleration gyroscope sensor (VG350) is placed into a belt worn at the waist to collect real-time data. Use of acceleration SVM ( Signal Vector Magnitude), angle, angular acceleration and displacement recognized five gestures (sit, squat, stoop, walk and run) by analyzing collected data. The experimental results show that the algorithm is simple and posture recognition rate is up to 96. 5%.

To overcome the problem which limited ability to suppress strength of PAPR(Peak To Average Power Ratio) of IPTS for optical OFDM(Orthogonal Frequency Division Multiplexing), an improved method divides the transmitting datas into D levels on the method of gradient descent, filtrating the trasmit datas layer after layer and
combining searching phase factors of each level with the IPTS technique. The simulation results show that the improved method is less complex than IPTS algorithm, and has lower PAPR than IPTS algorithm. In the 128 subcarrier OFDM system, the number of subcarriers packets V = 4, divided layer D = 2, and with the CCDF(Complementary Cumulative Distribution Function ) of 10-5, the PAPR in the improved algorithm is nearly lower 0. 5 dB than that in IPTS.

To solve the shortcomings that the Rule-based Channel allocation algorithm does not fit the practical application, we propose an algorithm considering the channel capacity and assigns the channel which has the maximum throughput to the node. The simulation results show that the proposed algorithm has greatly enhanced
throughput and fairness, and the complexity does not increase a lot. When the interference distance of 100 m and the number of cognitive nodes is 80, the proposed algorithm has 133% increase in throughput and 96% in fairness. The proposed algorithm can meet the actual needs better.

The sea clutter is affected by many factors, and is more complicated. The sea clutter noise has great impact on the estimated performance of the target signal's DOA(Direction Of Arrival) of Radar system. A new quaternion model to estimate signal's DOA combining with existing fractional lower order statistics algorithm
based on uniform electromagnetic vector sensor array is proposed. First, the electromagnetic vector array of quaternion was deduced, then the quaternion fractional low-order covariance was got by using fractional lower-order statistics. Finally, the quaternion field sparial spectrum was presented. The simulation results indicate that the method has a better performance than others in noise suppression and accuracy of the signal parameter estimation. Simulation results has demonstrated the effectiveness of the method.

Addressing the problems that the existing features of voice activity detection is susceptible to interference of various environmental noise, this paper proposes an algorithm using PM(Phase-Modulation) based feature. The feature of phase modulation is capable of fully characterizing the speech dynamic. And compared with the static features, it can more efficiently presents the difference between speech and noise to a higher degree which is the key for the well performed detection. The experimental results of voice activity detection using phase modulation feature and MFCC(Mel Frequency Cepstrum Coefficient) show that phase modulation feature is obviously better than MFCC.

In order to solve the problem of visual fatigue induced by parallax change discontinuity, a method of adjusting uncomfortable frame is proposed. Based on the comfort evaluation model, the motion and disparity vectors of the current frame derived from the JMVC(Joint Multi-view Video Coding) multi-viewing coding system are applied to find the video frame which can cause visual fatigue, and such frame is further adjusted using disparity adjustment of a uncomfortable frame to get a new frame which can enhance the visual comfort for eyes. Viewers make subjective scores according to their physical symptoms and fatigue response. According to comparisons before and after adjustment, the results show that the comfort level using the proposed method increases by 5.13% than prior through parallax adjustment for uncomfortable frames. The proposed method improves the comfort of viewing stereoscopic video, guaranteeing the quality of video.

The traditional direct torque control of induction motor has a disadvantage. The ripple of flux and torque get increased when the system run at a low speed. After analyzing the reasons, a novel direct torque control of induction motor using sliding mode control and SVPWM(Space Voltage Pwlse-Width Modulation) is
presented. The sliding mode controller contributes to the robustness of system by reducing the influence of disturbances and uncertainties. The space voltage PWM(Pwlse-Width Modulation) modulation is used to produce the desired space voltage vector and decrease the ripple of flux. We use speed sensorless technology to improve in dustrial applicability of the system. In Matlab/ Simulink circumstance, the results of simulation prove that the novel direct torque control system reduces the ripple of flux and torque effectively and gets accurate speed.

In order to estimate the vehicle yaw rate and increase the estimator accuracy, a method of BP(Back Propagation) neural network is adopted to estimate the vehicle yaw rate during the steering condition. There are four kinds of roads that are existing in reality: dry road, pitch road, watered road and ice road, and one neural
network cannot include the condition of four kinds of different roads. To solve the problem and increase the precision of the network estimator, we have trained four neural networks respectively to form a network group. By adding a selecting module to the system, the estimation value of yaw rate with corresponding road friction
coefficient can be picked out. We obtain the residuals of networks by co-simulation of AMESim and MATLAB.Finally, we evaluate and analyse the results of the estimation generating by the estimator. The method adopted is the data-based approach. Compared with the existing model-based method, it is independent of precise model and it can estimate the yaw rate precisely. Simulation results and analysis verified the viability and the precision of using BP neural networks to estimate vehicle yaw rate.

In a real scene, due to different illuminant color, corresponding points that are projections of the same scene point may have different RGB values of pixels and performance of stereo matching algorithms can be degraded. In order to improve the performance of algorithms, RGB proportional space is proposed based on the color formation model and pixel values are transformed into this space. We present an adapt support-weight algorithm based on RGB proportional space. Experimental results show that compared with the algorithm in the traditional RGB vector space, our algorithm can obtain better disparity maps.

为解决地面运动基座的轨迹跟踪控制问题, 基于其动力学模型, 利用动态反馈线性化技术实现了系统的完全状态反馈线性化。根据线性系统理论设计轨迹跟踪控制器, 并采取时变控制器增益的方法在满足控制力矩饱和约束的同时取得了良好的跟踪效果。Matlab 仿真结果显示, 采用该方法设计的轨迹跟踪控制器可使系统跟踪期望轨迹, 采取时变控制器增益的方法可使系统在跟踪过程中既满足了控制力矩饱和约束, 取得了良好的跟踪性能。与其他设计方法相比, 该方法能很好地将控制力矩饱和的情况考虑在内, 控制器设计简单且具有良好的稳定性。

Traditional active safety system‘s man-machine interaction function has poor performance, to solve this problem, we provide the system which is based on Freescale MC9S08DZ60 MCU. It provides dual execution modes, specifically, it can process the data from LIDAR and control the actuators independently. It can also upload the data to a host computer through a USB cable or WiFi signal as a gateway. It also allows the operator to configure thevigilance parameters of the system remotely. This system basically solved the problem of the low man-machine interaction and increased the interactivity between users and the active safety system.

In order to solve the shortcoming of fewer hand vein characteristics and lower recognition rates of hand vein recognition, we proposed a fusion method based on feature point distance of hand vein. Firstly, the hand vein is refined and matched the sum of quantized values of moment invariants feature; then the endpoints and the intersections of hand vein is extracted, and the improved Hausdorff distance is used to match, so the effects of zero point was removed, the degree of dissimilarity between two images was increased; finally, the two methods are weighted and fused. The experimental results show that fusion algorithm recognition rate can reach 96. 75%, and the running time is 0. 97 s.

Gait planning based on passive walking on even plane is presented for the four-link biped robot. The slope invariance control method is used to obtain the dynamics model of the robot in the horizontal plane. Then constant time scaling method is applied to scale the reference trajectory in the plane, so the walking gait whose stride length is constant and period is variable is got. Finally, numerical simulation is verified by Matlab, the results show that the method is feasible and effective.

In order to solve the noise information, which the leaf nodes votes in Hough Forest, this paper proposes an target tracking algorithm of the discriminative leaf based on Hough Forest. In the training phase of the original Hough Forest, leaf nodes do not have the ability to classify whether the image patches are target or background. The image patches contain a lot of noise, and aggregating voting information from all image patches extracted at each pixel location is really time-consuming. The experimental results demonstrate that the approach can reduce noise votes and increase tracking efficiency.

The finite time coordinated tracking problem is studied for multi-agent networks with inherent nonlinear dynamics under the condition that each agent has limited sensing range. Based on state information between agents, the distributed nonsmooth control protocol is designed that is capable of preserving topological connection assuming that nonlinear dynamics are satisfied with Lipschitz condition and initial network topology is undirectly connected. Sufficient conditions are developed to achieve finite time tracking by applying nonsmooth stability analysis methods, that is finite time tracking with topological connection can be achieved with suitable control gain. Simulation results verify the effectiveness and correctness of the proposed control scheme.

In order to improve the intelligence of ITS(Intelligent Traffic System), abnormal vehicles behaviors detection by means of the vehicles trajectory is proposed. The process is divided into three steps: target detection, tracking, vehicles trajectory analysis and vehicles behavior analysis. Firstly, the three frame difference method is used to achieve initially target location and the improved tracking algorithm based on Kalman predictor is used to track vehicle; then, the adaptive segmentation linear fitting algorithm based on the least squares method is used to achieve vehicle trajectory fitting quickly; finally, two parameters including the rate of velocity variation and the rate of direction variation are used to establish vehicle abnormal behavior detection model. Experiment result shows that the three high dangerous vehicle behaviors in the road surveillance videos can be detected quickly and effectively by the algorithms: sharp brake, sharp turn, and sharp turn brake.

For the shortcoming of guided filtering, such as slow operational speed and non-real time processing,the algorithm is speeded up based on CUDA(Compute Unified Device Architecture). In the proposed method,the sum of neighbor pixels-value is calculated based on CUDA parallel programming, and the mean value is calculated. The key parameters of guided filtering are obtained by taking advantage of texture memory and registers and algorithm optimizing. The whole optimum of approach is achieved. Experimental results show that, compared with CPU-based guided filtering algorithm, the proposed CUDA-based algorithm is greatly speeded up and basically meets the requirement of real-time processing.

In order to get different size of fluorescent semiconducting polymer nanoparticles, we adopted co-precipitation methods, used poly[(9,9-dioctylfluorenyl-2,7-diyl)-co-(1,4-benzo-{2,1,3}-thiadiaole)](PFBT) as original material. The obtained PFBT nanoparticles were characterized by DLS (Dynamic Light Scattering), TEM(Transmission Electron Microscopy), absorption and emission spectroscopy. The results show that the purity of solvent affects the conformation of polymer chain, and then affects the sizes of Pdots. The absorption spectra shows red shift, while the emission spectra presents blue shift with the particles size increasing.

In most of grassroots units for military base,all kinds of field communication equipment cannot adapt to the wild environment under the condition of the communication needs, because of their low efficiency in actual application, low utilization rate, portability and reliability is bad. In order to meet this demand. The uniqueness information acquisition terminal is designed based on the Linux platform S3C6410. It combined embedded processors, wireless networking, GPS(Global Positioning System) module, camera and RFID(Ratio Frequency Identification) module, collected the geographiclocation information, identity authentication information and image information, which integrated the uniqueness of data information display on LCD. The data information can transmit in real-time with fast connection to WiFi(Wireless Fidelity). After programming and debugging, it was proved to be stable, the uniqueness of data information transmission speed is excellent, the information is complete and reliable. It achieved the desired effect.

In order to achieve the micro four-rotor remote video transmission and attitude control functions, we design a remote monitoring terminal based on S3C6410+linux system. The main control unit of this terminal is S3C6410 ARM11 processor and the entire system is running on Linux operating system. The video and control information is transported via WiFi and we use UDP(User Datagram Protocol) communications protocol to transmit the information. The experiment results show that the terminal is stabile and the transfer rate can satisfy the needs of real-time monitoring.

According to the problems exist in the PC platform of large volume, high-power, high cost and poor portability, the author design an embedded terminal for secure authentication based on facial features. I choose ARM11 development board as the hardware platform, Linux operating system as the software platform, and use transplant Qt graphical interface library and Opencv computer vision library on the platform. Face detection was achieved by AdaBoost algorithm, face recognition was achieved by PCA ( Principal Component Analysis) algorithm. The test results show that this terminal can realize the function of face recognition, recognition accuracy is high and recognition time is short, the desired goal is achieved.

Automatic berthing system is a kind of environment through the information collection around to find suitable berthing vessels, automatically enter the position control system. According to the automatic berthing method time-consuming problem in foreign and domestic, there is an urgent need to develop a stable system effectively realizing independent berthing. Combined with the actual of berthing motion model analysis, a parallel, oblique three berthing fuzzy controller is designed whose the simulation and verification were performed in the Matlab/ Simulink. According to the proportion of each component of ships, a model is designed to make automatic berthing test by using ultrasonic distance measurement to obtain the surrounding environment data technology and converting to C code. The test results show that the system can effectively realize independent berthing, and provide a theoretical reference value for the further research and development.

Because of the influence of illumimation changes and insurfficient light, license plate location is much more difficult and has lower accuracy. The phase consistency was used for extracting the image edge information, and character texture was used for extracting the plate area in image. Using vertical projection to get the one-dimensional signal of scanning plate area. The interference of noise is filted out by using the wavelet multi-resolution characteristics. Finally, using statistic characteristics of license plate to extracte it. The simulation results show that the the accuracy of this algorithm can achieve 96% in the illumination and low light environment.

To solve the problem that hologram contains a large amount of data and performance of reproduced hologram is not ideal. A method is proposed to adjust dynamically learning factor and inertia weight for the PSO (Particle Swarm Optimization). The MPSO-BP(Modified Particle Swarm Optimizing BP Neural Network) is constructed combining the modified algorithm with BP(Back Propagation) neural network. Comparing with the BP neural network compression and PSO-BP(Particle Swarm Optimizing BP Neural Network) compression, the network has the advantages of better performance of reproducing hologram while maintaining a good compression efficiency.

为了消除大气湍流对图像的影响, 提高图像质量, 结合稀疏表示理论, 采用字典学习的算法处理大气湍流退化图像。将DCT 过完备字典、K-svd 全局字典和自适应字典的算法应用于图像去噪过程, 并与维纳滤波算法进行比较。结果表明, 该算法能较好地滤除大气湍流退化图像的噪声, 提高图像的峰值信噪比。仿真实验验证了稀疏表示在处理大气湍流退化图像的可行性, 对比传统算法具有更好的去噪性能。