Research On Human Gait-based Identification System

Gait Recognition is one new and emergent research topic in the field of biometric recognition, which makes use of human walking pattern to recognize and identify, especially for human identification at a distance in non-contact situation. Now, it has received growing interest within the pattern recognition community and become one hot research area in the field of biometric research.In this paper, human identification technique based on human gait feature is researched and it mainly following problems such as human motion detection and tracking, gait feature extraction, gait recognition in single view and multiple views gait information fusion. Firstly, two background modeling methods such as median and Gaussian modeling are proposed and human motion is detected using background differencing. Secondly, one color calculation model is adopted to compute the chromatic aberration to eliminate the shadow and the noise can be removed through dilation and erode operators to obtain human motion region based on connected component analysis. Thirdly, human gait silhouettes are obtained through counter tracking and the infection that small hole in human regions is avoided. Then, three gait feature method based human silhouette analysis such as calculating the distance between the pixel in the silhouette and its centroid, Fourier descriptors and wavelet descriptors. Fourthly, two recognition algorithms such as principal component analysis (PCA) and nearest neighbor (NN) classifier and independent component analysis (ICA) and support vector machine (SVM) are proposed in single view separately. They describe gait sequence as one whole objective and compress its feature using PCA and ICA and receive comparatively high recognition rate. Lastly, one fusion method is proposed based on the fact that one gait sequence takes difference feature from different multiple views. It fuses the different gait feature from multiple views and finish human identification through multiple gait information fusion. The method is tested in two gait databases and receives encouraging recognition accuracy and the results indicate that gait recognition using multiple views information fusion can obtain higher recognition rate than any other single view.Meanwhile, one human identification system which provides one testing platform algorithm is designed and the conclusion that gait can be applied to human identification at a distance is made.