算法提出了

一个新的部位观测模型和一种新的减小部位状态空间的方法：（1）对人体不同部位采用不同尺寸的细胞单元计算HOG特

征，并利用线性SVM进行分类，从而提出一种新的部位观测模型；（2）利用人体部位定位的先验分布确定部位定位区域，然

后通过邻域归并和设置与部位模板的匹配度阈值进一步减小状态空间，从而提出了一种减小部位状态空间的方法。仿真

实验结果表明所提算法与传统算法相比更加有效。-Algorithm proposed site of a new observation model and a new state-space methods to reduce parts: (1) to different parts of the body cells of different size unit calculates HOG features and linear SVM classification, which proposes a The new site observation model　(2) the use of body parts to determine the positioning of the prior distribution site positioning region and then further reduce the state space by matching threshold neighborhood merging parts of the template and settings, thereby reducing the site presents a The method of the state space. Simulation results show that the algorithm is more efficient compared with the conventional algorithms.