AGRS: Automated gait recognition system in smart environment

摘要

Due to advancement in technology, surveillance systems have become more automated without manual assistance. This paper presents a new method of gait feature extraction and fusion algorithm to identify the individuals irrespective of the walking style of the person in various occlusion states. This is done in addition to their walking speeds and varying clothing style. The Gait Recognition is performed based on the bio mechanics approach. The kinematics features are also called as dynamic features such as joint angles are extracted from the video sequences. The novelty of this paper lies in the extraction of static and dynamic features and feature fusion methodologies, which involves model-based parameters. The static features are measured from the distance function and the dynamic features such as joint angles are extracted from the transformation matrix. The fusion of both features forms a final feature vector for every person to reveal the identity of person irrespectively based on various factors such as occlusion state: static and dynamic occlusion, normal, fast and slow walking speeds and different clothing. The system has been analyzed using CMU motion capture dataset, TUMIIT KGP database and real time videos for person identification by various factors. This smart system can be used in apartments to identify the entry of unauthorized people and to avoid theft and burglary cases.