点击次数：

论文类型：期刊论文

发表时间：2011-12-01

发表刊物：Yi Qi Yi Biao Xue Bao/Chinese Journal of Scientific Instrument

收录刊物：EI、PKU、ISTIC、Scopus

卷号：32

期号：12 SUPPL.

页面范围：143-146

ISSN号：02543087

摘要：In this paper, a algorithm based on shape context and polar coordinate histogram is proposed to recognize human actions using difference silhouettes obtained from action silhouette Sequences. Difference silhouettes are extracted from binary silhouettes of different frames as action features, and contours and samples could be collected. Polar histogram and shape context are used for filtering some dissimilar action and shape matching. In the framework of the nearest neighbor method, the improved average Hausdorff distance is proposed as the similarity index for action recognition. Simulation results show that action recognition based on the proposed method has higher classification accuracy.