Using Gait to Recognize People

In modern society, the need for the establishment of defense and prevention mechanisms has encouraged the development of automatic human recognition systems based on biometrics, i.e. the analysis of a person´s psychological and behavioral features. Gait, or the peculiar manner of walking, allows the recognition process to be made at a distance since it is possible to extract the gait information from a video sequence of a distant person walking. This paper proposes a gait recognition algorithm based on the averaged silhouette of a person over a gait cycle. A binary silhouette of the walking person is obtained by background subtraction; the binary silhouettes are then aligned and averaged over each gait period. The Euclidean distance between the averaged silhouettes of a number of persons is used for recognition purposes. Experimental results, using both lateral and oblique views, show very promising recognition rates