Human Gait Classification Using Motion Information

Human Gait Classification Using Motion Information

10.6302/JITA.200605_1(1).0001

Ming-Gang Wen；Chin-Chuan Han；Chin-Chen Chang；Cheng-Hsiang Liu

video surveillance ； moving target detection ； target tracking ； optical flow ； hidden Markov model

Journal of Information Technology and Applications(資訊科技與應用期刊)

1卷1期（2006 / 05 / 01）

1 - 8

英文

In this paper, human gaits in the video streams were identified using the local motion features and the hidden Morkov model (HMM) method. First of all, the regions of multiple moving targets, called region of interest (ROI) are detected, labeled, and tracked from the image sequences in the clustering background. The local motion vectors relative to the ROI's center are used to represent human activities. These vectors are transferred to the sequence of states to form the training and testing samples using the clustering algorithms. Four activities, including walking, running, hopping, and limping, are specified and designed for identifying human activities in video streams. Some experimental results are illustrated to show the validity of the proposed methods. Finally, the conclusions and future works are given.