Gait Recognition using Time-of-Flight Sensor

This paper develops a biometric gait recognition system based on 3D video acquired by a Time-of-Flight (ToF) sensor providing depth and intensity frames. A first step of the proposed gait analysis is the automatic extraction of the silhouette of the person via segmentation. The segmentation of the silhouette is performed on the depth frame which provide information which describes the distance from the camera of every pixel in the intensity frame. The range data is sensitive to noise thus we apply morphological filtering operations to enhance the segmented object and eliminate the background noise. The positions of the joint angles are estimated based on the splitting of the silhouette into several body segments, based on anatomical knowledge, and ellipse fitting. The resulting parameters from this analysis of the silhouette are used for feature extraction from each frame. The evolutions of these features in time are used to characterise the gait patterns of the test subjects. Finally, we do biometric performance evaluation for the whole system. To the best of our knowledge, this article is the first article that introduces biometric gait recognition based on ToF Sensor.