The Research Of Gait Recognition Based On Feature Fusion

Gait recognition which is using the posture of the people on the identification of the people is a new biometric technology. This technology is non-invasive, long-distance identification and the biological characteristics is difficult to hide which often occurs in other technology such as fingerprint characteristics,iris.Therefore,the gait identification as a pattern recognition emerging sub-fields. more and more researchers in this field of research take the attention, and in real life it has been widely used in the video,especially widely in security and defense and the monitoring. Generally speaking, the gait recognition process is divided into three parts: gait extraction, the feature extraction and gait recognition. Comprehensive literature, extracting gait features and classification algorithm has been highlight in the field of pattern recognition. Although the research of the gait made certain progress,because of the diversity of the gait and the complexity of the gait surroundings, there is not a general recognition algorithm. Therefore, this paper conducts from extraction gait characteristics and gait algorithms of some key problems this field,.The main research contents include the following aspects:1 The paper focus on describing the various methods of gait periodic, presents a new method of describing the gait periodically from the source image. This method focus on the part below the hip of the object and extracts the center of the hip from the image, including a state gait starting from this paper (maximum span for initial gait analysis) to a similar to the end, and then get a gait cycle of image sequence.Experimental results show that this method is not only robust, but also highlights a important properties of gait periodic.2 The thesis which combines the physiological structure characteristics and the motion characteristics combines the static characteristic and dynamic characteristics of gait. According to the characteristic of gait movement, the thesis extracts the motion features of key points of the object parts below hip,knee,including speed-torque and distance-torque. The speed-torque describes the trajectories of gait and the distance-torque describes the consistency of the sport; Combined with the physical structure characteristics of the distance from the knee to the ankle, the thesis extracts the features for gait. Then the static and dynamic characteristic are fused for gait recognition. Using various classification algorithm to recognize gait, the experiment results show that the feature is effective and can be used to improve the gait recognition rate.Experimental results verify that using centroid of the following section below hip joint describes gait periodic has more advantages and using dynamic characteristic combined with static characteristic is feasible. Although we have done some work in gait feature extraction, further study is necessary.