Gait recognition using partial silhouette-based approach

Silhouette-based gait analysis is a well-established biometric approach for human identification. Over the years researchers have proposed a number of gait recognition approaches based on the entire silhouette of human body. These approaches are proven to give good recognition accuracies. However, the feature vector generation and subsequent classification depend on information extracted from the whole object silhouette involves handling of considerably large data size. In this paper, the authors propose a new method for human identification considering the fact that the partial silhouette of a human body often contains sufficient discriminating information for gait recognition. The idea is based on extracting features from the portions of the silhouette that contains one of the most dynamic features of gait - the swinging hands of a human being. The proposed method is tested using two standard, widely-used public gait datasets. Results show the effectiveness of the proposed methodology.