Research And Implementation Of Gait Recognition System Based On Contour Features Double Support Phase

As mankind enters the information age, and people pay attention to security issues, moreand more surveillance video storage in digital video format. Become very practical significanceto recognize the video information of the identity-based. Gait recognition is an emergingbiometric identification technology, is designed by people walking posture identification,compared to other biometric technologies, gait recognition has a non-contact distance and noteasy to disguise advantages. It features very consistent application needs intelligent monitoringfield, so gait recognition is becoming a hot topic today.Themselves very easy gait recognition, beyond a certain distance can distinguish peoplewho know people based on walking gait. But the amount of data gait video is very large, so thecalculation is very large amount of gait recognition, treatment is also more complicated anddifficult. Although the biology and mechanics of the human gait done a lot of research work,but based on studies of human gait identification is still in the experimental stage. So far, nocommercial gait-based identification system.In this paper, based on the human side gait video gait recognition system to achievepractical. By analyzing a large number of gait characteristics, as well as the aspect ratio of therectangular outline of the tilt angle of the thigh and calf of the gait cycle showed a cyclicalchanges. The outline of the rectangle aspect ratio and tilt angle as the main gait. First, accordingto the characteristics of video images of moving objects, moving target detection and tracking,extract the movement area of an image using background subtraction method based onimproved. Then noising by morphological processing, extracting human gait is a rectangularoutline when the aspect ratio of the double support phase as gait; thigh and straight legsilhouette topical application of Hough transform to detect and then the double support phasetime, resulting in the thigh and tilt angle as another leg gait. Finally, the use of hidden Markovmodel gait feature extraction algorithms to classify training.In this paper gait recognition system based on the OpenCV vision library implementation.Experimental samples are actually collected multiplayer side gait video library. Experimentalresults show that this study video-based gait recognition scheme very practical, in the case ofgait database close to real life also get a better recognition rate.