Gait Recognition Based On Depth Information Gait Recognition Based On Depth Information

Abstract:Gait recognition is a new method of biometric identification, it is different from other ripe methods of biometric identification such as fingerprint identification, iris recognition and face recognition and so on, gait recognition is still just in researching by researchers. However, compared with other methods, gait recognition has the following advantages:firstly, it is more accessible for people needn’t touch the device when it works; secondly, it doesn’t call for high quality images; lastly, people’s gait can’t be changed or forged easily in a period. With these advantages, gait recognition could be used in long distance or large scale place such as some public places or important places, and discover the potential danger in advance. Because of the widely application of gait recognition, more and more research institute from all over the world are taking up the research.Usually, gait recognition is researched based on plane images or3D body information. This paper is based on3D body information. Currently, most researches which based on3D body information obtain the3D body information with several cameras working at the same time. As using several cameras is more expensive and lacked of practical, this paper use only one kinect to get3D body information, and we propose a series of methods to simplify and compress the body point cloud. In this paper, we propose two new gait representation which show a better result than others. The main work of this paper is listed as follows:1. Processing the original body point cloud sequences. The processing procedure includes normalizing body point cloud, determining the gait cycle, selecting key frames, resampling the body point cloud and building depth gait feature images.2. Building DGEI and DGGEI, which are the new gait representations. Doing experiments with GEI, DGEI and DGGEI. Comparing and analysing the results of the different experiments.3. Developing a prototype system with Visual Studio2008. The system is a integration of the method in this paper, and can be used in researching and demonstration.