点击次数：

论文类型：会议论文

发表时间：2014-01-01

收录刊物：EI、CPCI-S、SCIE、Scopus

卷号：8866

页面范围：489-498

关键字：Image representation; Action classification; Difference silhouette; Static reservoir; gamma - C plane

摘要：In this paper, the variation between features of frames for human action recognition is studied, and a new local descriptor extracted among the differences of human silhouettes is posed. This descriptor is represented by coarse histograms based on the distribution of sample points on the outlines of difference silhouettes. The static reservoir is employed as the classifier of human action. Two hyper- parameters, the scaling parameter gamma descriptor and the regularization parameter C are taken to characterize a static reservoir, and the proper static reservoir for action recognition is identified on the gamma - C plane. We test our approach on two commonly used action datasets, and the experimental results show that the proposed method is effective.