A Gait Classification System using Optical Flow Features

Gait classification is an effective and non-intrusive method for human identification. This paper proposes a system to recognize human identity using optical flow features. The distinguishing characteristic of the proposed system is that we only adopt optical flow information and do not consider shape features or other information. The moving object is detected and located from the flow field using a gaussian model. Afterwards, each subject is identified via the established histogram using optical flow features. The proposed system applies and compares three different kinds of optical flow extraction algorithms. Various experiments with two different databases analyzed and discussed the feasibility of the approach. This work demonstrates that optical flow information is useful for gait classification even for unstable optical flows.