A Novel Human Gait Recognition Method by Segmenting and Extracting the Region Variance Feature

Existing methods of gait recognition suffer from some shortcomings, which are discussed at the beginning of the full paper. In order to suppress these shortcomings as much as possible, we proposed a new automatic gait recognition approach based on the region variance feature. Firstly, the binary silhouette of a walking person is detected from each frame of the monocular image sequences. Then we divide the two dimensional silhouette of the walker into three regions (head region, trunk region and legs region). Next, the variance features of these regions are extracted respectively. Together with the ratio of the silhouette´s height and width, the gait signature vectors are constructed to identify different subjects. Finally, similarity measurement based on the gait cycles and NN and KNN classifiers are carried out to recognize the different subjects. Experimental results show that the proposed novel method is very effective and correct recognition rates are over 92% and 97% on UCSD and CMU database, respectively