Real-time recognition of humans by their walk.

In this paper, we will present a novel method of human recognition using gait information. The position of the moving object is first detected using the difference between two continuous frames. Then, based on the detected moving object, all of the gait information is exploited. Based on the head, shoulders and back contour, the walking direction can be extracted. Depending on this parameter, the lateral or frontal characteristic is focused on in the features detection process. Once all of these characteristics have been exploited, a feature vector is constructed for use in the recognition process. To evaluate the effect of the proposed method and compare it with other methods, we present some simulation results obtained in both indoor and outdoor environments.