Gait Recognition based on Dynamic Texture descriptors

The human movement analysis is an attractive topic in biometric research. Recent studies indicate that people have considerable ability to recognize others by their natural walking. Therefore، gait recognition has obtained great interest in biometric systems. The common biometrics is usually time-consuming، limited and collaborative. These drawbacks pose major challenges to the recognition process. Gait analysis is inconspicuous، needs no contact، is difficult to hide and can be evaluated at distance. This paper presents a bag of word method for gait recognition based on dynamic textures. Dynamic textures combine appearance and motion information. Since human walking has statistical variations in both spatial and temporal space، it can be described with dynamic texture features. To obtain these features، we extract spatiotemporal interest points and describe them by a dynamic texture descriptor. Afterwards، the hierarchical K-means as a clustering algorithm is applied to obtain the visual dictionary of video-words. As a result، human walking is represented as a histogram of video-words occurrences. The performance of our method is evaluated on two dataset: the KTH and IXMAS multiview datasets.