Identification of a Person by Gait on the Basis of Wavelet Technology

A new method of gait recognition based on a sequence of frames from a video sequence is proposed. The silhouette of a moving figure is extracted by subtracting the background from the image. The contour of the silhouette is considered as a one-dimensional signal in the polar coordinate system. Signal values are the distances from the center of gravity of the silhouette to the points of the contour, and the argument is the angular coordinate. From a set of one-dimensional signals for successive frames, a two-dimensional signal is formed, which is transformed into a grayscale image. The representation of the gait fragment in the form of an image allows using the latest research in the field of intellectual processing of graphic data for gait analysis.In this paper, the Gabor wavelet method is used to identify the person by gait. A program was developed for Visual Studio using the OpenCV graphics library. Testing the program on samples from the CASIA gait database showed high recognition accuracy.