Gait recognition based on key frame and elliptical model

In this paper, a robust gait recognition algorithm based on key frame and ellipse model is proposed, which solves the problems including finding key frames on discontinuous contour segmentation and accurate feature extraction on ellipse model. Firstly, a robust extraction algorithm of key fame is introduced, which makes detection of key fames easier and more accurate. Secondly, human bodies corresponding to nearest and furthest distance between two legs in the key frame are separately divided into four parts composed of head, torso, and lower limb above knee and under knee, and six parts consisting of head, torso, left and right calves, and left and right thighs. These parts are fitted to corresponding ellipse parameters by the new fitting algorithm. Thirdly, an improved algorithm based on ellipse fitting is proposed in the paper, which calculates accurate ellipse parameters as human feature. Lastly, according to accurate parameters on ellipses, Euclidean and Mahalanobis distances are used to judge human identification. Experimental results show that recognition algorithm is effective.