Human Gait Recognition
project report tiger|
Active In SP
Joined: Feb 2010
08-02-2010, 10:41 AM
Human Gait Recognition.pdf (Size: 452.87 KB / Downloads: 169)
The reliable extraction of characteristic gait features from image sequences and their recognition are two important issues in gait recognition. In this paper, we propose a novel 2-step, model-based approach to gait recognition by employing a 5-link biped locomotion human model. We first extract the gait features from image sequences using the Metropolis-Hasting method. Hidden Markov Models are then trained based on the frequencies of these feature trajectories, from which recognition is performed. As it is entirely based on human gait, our approach is robust to different type of clothes the subjects wear. The model-based gait feature extraction step is insensitive to noise, cluttered background or even moving background. Furthermore, this approach also minimizes the size of the data required for recognition compared to model-free algorithms. We applied our method to both the USF Gait Challenge data-set and CMU MoBo data-set, and achieved recognition rate of 61% and 96%, respectively. The results suggest that the recognition rate is significantly limited by the distance of the subject to the camera. 1.
1Rong Zhang 2Christian Vogler 1Dimitris Metaxas
1 Department of Computer Science 2 Gallaudet Research Institute
Rutgers University Gallaudet University
Human recognition is an important task in a variety of applications, such as access control, surveillance, etc. To distinguish different persons by the manner they walk is a natural task people perform everyday. Psychological studies [10, 19] have showed that gait signatures obtained from video can be used as a reliable cue to identify individuals. These findings inspired researchers in computer vision to extract potential gait signatures from images to identify people. It is challenging, however, to find idiosyncratic gait features in marker-less motion sequences, where the use of markers is avoided because it is intrusive and not suitable in general gait recognition settings. Ideally, the recognition features extracted from images should be invariant to factors other than gait, such as color, texture, or type of clothing. In most gait recognition approaches [6, 16, 11], recognition features are extracted from silhouette images. Although these features are invariant to texture and color, the static human shape, which is easy to be concealed, inevitably mingles with the movement features In this paper, we propose a 2-step, model-based approach, in which reliable gait features are extracted by fitting a five-link biped human locomotion model for each image to avoid shape information, followed by recognition using Hidden Markov Models (HMMs) based on the frequency components of the trajectories of the relative joint positions. Applying our approach to both the USF Gait Challenge data-set and the CMU MoBo data-set, we demonstrate that promising recognition rate can be obtained using gait only features.