为了解决人的衣着变化和携带物品对步态识别的影响，提出了一种基于动态部位变化的步态识别方法。首先应用背景差分和阴影消除获得人体步态轮廓，并对获取的轮廓进行位置中心化和大小归一化；然后通过步态能量图和阈值分割的方法划分出每一帧的动态部位，并使用扇形区域距离变换的方法对动态部位进行特征提取；最后使用最大熵马尔可夫模型对各个人的步态进行建模，完成了基于概率图的识别。该方法在CASIA步态数据库上进行了实验，取得了较高的正确识别率，实验结果表明该方法对人的衣着变化和携带物品情况下的步态识别具有较强的鲁棒性。

To solve the problem that the clothes variance or taking goods may affect the result of gait recognition, a new gait recognition method based on variance of dynamic region is proposed in this paper. Firstly, through the background subtraction and shadow elimination, human motion silhouettes are obtained, which will be normalized in terms of location and scale. Next, the dynamic regions are obtained using gait energy image and threshold segmentation, and gait feature is extracted from the dynamic region using the sector region distance transform. At last, maximum entropy markov Model is used to model the gait sequences of each people and implements recognition based on probability graph. The method is evaluated for the CASIA gait database and receives comparatively high correct recognition rate. The experimental results show that our approach is robust in the case of clothes variance and taking goods.