Human activity recognition using standard descriptors of MPEG CDVS

In this paper, a novel human activity recognition method is presented. The proposed solution uses a single stationary camera in order to detect common human activities like: hand waving, walking, running etc. Unlike other methods which use different kinds of characteristic point descriptors in order to describe human poses, the proposed solution uses a CDVS descriptor which is part of the MPEG-7 standard. This allows the efficient calculation of a compact descriptor in a camera. The main goal of this work is to propose an efficient application of CDVS to describe and recognize different human activities. The experiments were performed on the Weizmann and KTH datasets. The obtained results prove the high accuracy of human activity recognition system with CDVS.