摘要： 为在降低样本训练时间的同时提高数据分类能力，提出一种基于加权行质量的步态识别方法，提取人体轮廓行质量向量作为步态特征，分析特征向量各元素的贡献度从而对特征向量进行加权，采用归一化欧氏距离度量相似度，并使用最近邻分类器进行分类。在CASIA数据库上的实验结果表明，该步态识别方法既满足步态识别对实时性的要求又保证较高的识别率。

关键词: 步态识别, 线性判别分析, 行质量向量, 归一化欧氏距离

Abstract: In order to reduce the training time while increasing the sample data classification capabilities, a new approach to the problem of gait recognition based on weighted row mass vector is proposed, which uses the row mass vector as gait feature, weights it by analyzing the contribution of each mass vector element. The Normalized Euclidean Distance(NED) is used for matching and the nearest neighbor classifier is used for classifying. Experimental results on CASIA gait database show that the proposed algorithm can greatly reduce the training time and achieve high recognition rate.

Key words: gait recognition, Linear Discriminant Analysis(LDA), row mass vector, Normalized Euclidean Distance(NED)

中图分类号: 

• TP391