中文摘要：

提出了一种基于肢体长度参数的用于多视角情况下的步态识别算法．利用脚间距计算方法和动态身体分割方法，拟合出场景的转换参数，并以此估计出人运动情况下的5个肢体长度参数，使用标准的分类器进行分类．在中科院自动化所提供的NLPR数据库上做了大量仿真实验，结果表明本方法在人的行走方向相对于摄像机光轴的角度改变时仍然保持了较高的识别率．

英文摘要：

A multi-view method based on the length of body＇s parts is presented for gait recognition. Through distance- between-feet extraction and dynamic body labeling, we estimated the converter parameters of the surveillance scene and estimated five static body length parameters to denote gait signature. Experiments made on the NLPR database show that this method is quite effective in distinguishing identity, and has a high recognition rate even when the walking people are changing directions.