A Model Based Approach for Gait Recognition System

In this paper, we propose a model based approach for gait recognition using the mathematical theory of geometry and image processing techniques. In such approach, feature matrices used for gait recognition are constructed using segmentation, Hough transform and corner detection techniques. Indeed, it is possible to recognize a subject by analyzing the gait parameters extracted from his footsteps taken in different frames. In the preprocessing stage, the picture frames taken from video sequences are inputted to Canny Edge detection algorithm in order to detect the image edges and to reduce the noise by means of Gaussian filtering. The Hough transform is then applied to isolate the features of the preprocessing output and to get a gait model. The latter is used to extract the gait parameters, and the Harris Corner Detection technique is used to detect the corners and to generate the feature points. The gait parameters are measured by means of feature points and then stored in a gait database. Using a gait recognition interface the random subjects parameters are compared against a template set in the available database for recognition. In the proposed method, we have considered a database including ten subjects and a five parameters based gait recognition system. It is worth noting to remark that when the camera is placed at 90 and 270 degrees towards the subject, all the recognition parameters are clearly visible, measurable and lead to have more than 80% accuracy in recognition results. Keywords— Biometric, Gait recognition, Canny Edge Detection, Hough Transform, Harris Corner Detection