Gait recognition based on leg motion and contour of silhouette

Human identification at a distance has recently gained growing interest from computer vision researchers. This paper presents an automatic gait recognition system that recognizes a person by the way they walk. The gait signature is obtained based on the angle and the contour of the silhouette. For each image sequence, background subtraction is used to extract moving silhouettes of the walker. The silhouette is divided using three parameters: the perimeter of the contour of the silhouette, the area of the silhouette and the angle between right leg and left leg. Principal component analysis (PCA) was then applied to this dataset in order to reduce the dimensionality. The perimeter of contour and the area of each silhouette was added to the dataset. This paper applies dynamic time warping to distinguish between the different gaits of human subjects. The performance of the proposed method is tested using CASIA database B. Recognition results show this approach is efficient.