基于特征步态的身份识别方法

Automatic Human Identification Based on Eigengait

本文提出了一种有效实用的基于步态的身份识别方法。通过提取目标行走的差分步态图像,将目标运动过程的动态信息转化为二维平面中的静态信息,利用K-L变换进行特征提取,使用标准的分类器进行分类。在中科院自动化所提供的NLPR数据库上做了大量的仿真实验,验证识别性能。

In this paper, we propose a new gait recognition method based on eigengait. In the method, differential gait images are introduced, in which dynamic information of walking can be expressed as static information in a 2-D plane. K-L transformation is employed to generate eigengait vectors. In recognition process, standard classifiers such as ENN and NN are used for classification. Extensive experiments have been made on the NLPR database and the results prove the higher performance of the method.