基于步态动态图的步态识别方法.PDFVIP

第10卷第16 期 Vol.10 No.16 2017 年 8 月 August 2017 基于步态动态图的步态识别方法 周 磊，叶俊勇* （重庆大学光电技术及系统教育部重点实验室，重庆 400044 ） 摘要：为有效地从多帧动态步态序列中提取步态特征并进行身份鉴别，采用一种基于步态动态图（gait dynamic image ，GDI ）的步态特征提取方法，主要通过排序池化（rank pooling ，RP ）方法对步态序列的动态信息进行 编码。将排序池化方法作用于步态序列，生成对应的步态动态图；然后使用核主成分分析（kernel principal component analysis ，KPCA ）方法提取步态动态图的主成分，其可以对图像像素间的相关性进行描述；最后将 主成分与泛化能力较强的支持向量机（support vector machine ，SVM ）分类器相结合进行步态识别。实验结果 表明，本文方法取得了较理想的识别效果。 关键词：图像处理；步态识别；步态动态图；排序池化；核主成分分析；支持向量机 中图分类号：TP391.4 文献标识码：A 文章编号：1674-2850(2017)16-1811-08 Gait recognition method based on gait dynamic image ZHOU Lei, YE Junyong (Key Laboratory of Optoelectronic Technology and Systems, Ministry of Education, Chongqing University, Chongqing 400044, China) Abstract: In order to extract gait features more efficiently from a sequence of multi-frame dynamic gaits and identity recognition, a novel approach to extract gait features based on gait dynamic image (GDI) was proposed, which encodes dynamics of a gait sequence by rank pooling (RP) method. Gait dynamic images were obtained by directly applying rank pooling on gait sequences. Kernel principal component analysis (KPCA) extracted principal components from gait dynamic images with nonlinear method, and described the relationship among three or more pixels of the identified images. The support vector machine (SVM) classifier is adopted based on its excellent generalization capacity in classification problem to gait recognition, combined with principal component analysis. Experimental results show that the method is capab