Gait refers to the style of walking of a person and is considered to be a potential biometric feature for identification. Although several approaches for fronto-parallel gait recognition using RGB cameras have been proposed over the last two decades, significant attention has not yet been given to frontal gait recognition. This is due to the fact that, shape variation of a walking subject cannot be effectively obtained from the binary silhouettes extracted from frontal view RGB sequences. Research on frontal gait recognition was benefited significantly after depth cameras like Kinect were made available commercially. Although, a few recently proposed methods use the RGB-D streams from Kinect for frontal gait recognition, none of these considers the presence of incomplete cycle sequences, or occluded gait sequences. It may be noted that, presence of occlusion or incomplete gait cycle information is almost inevitable in most real-world scenarios. My Ph.D. dissertation attempts to address these challenging problems using Kinect as the surveillance camera. In the talk, I will first highlight the main contributions of my Ph. D. work. Next, I will specifically focus on describing our approach towards performing recognition in the presence of incomplete gait cycle information. We have considered airport security check point as a typical application scenario, where two depth cameras mounted on top of a metal detector gate positioned beyond a yellow line, respectively, record front and back views of a subject as he goes through the check-in process. Due to the short distance of the surveillance zone between the yellow line and point of exit, it is often not possible to capture a full gait cycle independently from the front view or back view. An initial stage of anthropometric feature-based classification followed by motion feature extraction from the front view is used to restrict the potential set of matched subjects. A final classification is then applied on this reduced set of subjects using depth features extracted from the back view. We have performed extensive experiments using gait data from 60 subjects captured in an indoor environment. The proposed approach has been seen to perform robustly in the presence of glossy floor, in case of near-frontal walking direction, as well as, in the presence of multiple persons in the background. The method has a response time of about 9 seconds when two training sets for each subject are used, but shows a much higher recognition rate compared to the existing gait recognition approaches.
Dr. Pratik Chattopadhyay obtained the Ph.D. degree in November 2015 from the School of Information Technology, IIT Kharagpur, under the supervision of Prof. Shamik Sural. He PhD thesis title is Frontal Gait Recognition using RGB-D Cameras and his reasearch interests include: Image Processing, Machine Learning, Pattern Recognition with specific interest in Gait Recognition and Person De-identification