中文摘要：

提出了一种通过统计行走模式的动态信息进行步态识别方法．对代表每一类的步态能量图像（GEI）进行方差分析，以求得动态权值掩模（DWM）．通过DWM对原始GEI进行动态和形状信息的增强，以获得新的步态表征EGEI．为增加可辨识信息，使用一组Gabor小波对EGEI进行卷积，然后采用辨别共同向量分析（DCV）将高维卷积结果在低维空间表示．通过使用简单的分类策略在USF步态数据库上的对比实验，证明了本方法对识别性能提高的有效性．

英文摘要：

This paper presented a novel approach for gait recognition by analyzing the dynamics of walking patterns. Inspired by intuitive inference, a dynamic weight mask （DWM） is created by the variance of averaged gait energy images （GEIs） representing each class. A dynamics and shape enhanced gait representation called EGEI is obtained from DWM. To increase discriminative information an ensemble of Gabor kernels is convolved with EGEIs. Gabor gait is represented in the lower dimensional space using discriminative common vectors analysis. The proposed algorithm was tested on the USF HumanID Database. The experimental results demonstrate that each step in the proposed algorithm is effective and the recognition performance is improved.