Gait Recognition via Fused Hidden Markov Models

This paper presented a novel gait recognition approach based on Haar wavelet and fused Hidden Markov Models. It solves the problem that key points in each region represent gait feature insufficiently. Firstly, images from video sequences are converted into binary silhouette. Haar wavelet transform is employed to obtain key points for distinct features, and the key points are analyzed. Two sub images are utilized to represent gait features in each silhouette, and employ Principal Component Analysis to reduce its dimensionalities. Finally, fused Hidden Markov Models are employed to train and test, and it is helpful in analyzing features. Consequently, we can not only simplify the process, but also improve the recognition accuracy.