Frontal Gait Recognition Combining 2D and 3D Data
Soares, L. D.
Frontal Gait Recognition Combining 2D and 3D Data, Proc ACM Workshop on Multimedia and Security - MMSec, Conventry, United Kingdom, Vol. -, pp. - - -, September, 2012.
Digital Object Identifier:
Automatic human recognition systems based on biometrics are becoming increasingly popular in today’s world. In particular, gait recognition (i.e., recognition based on the peculiar manner of walking) is especially interesting since it allows the recognition even at a distance, without explicit user co-operation. In this paper, depth information is integrated in a silhouette-based gait recognition scheme, in order to produce a hybrid 2D-3D frontal gait recognition scheme. The depth information of the human silhouette is obtained with a Kinect camera, from which a three dimentional (3D) human point cloud is obtained. In the proposed multimodal framework, feature extraction is done by considering the full 3D human point cloud, as well as three 2D projections of it, corresponding to the top, front and side-views of the human subject. The experimental results for this multimodal scheme, based on both 3D and 2D data, show superior recognition rates, when compared to existing front-view gait recognition schemes.