Gait Recognition Based On Wavelet Descriptor And Human Skeleton Model Through Multiple Views Fusion

Gait recognition is a relatively new subfield in biometric recogtion, which attempts to recognize people from the way they walk or run. With a growing need for a full range of visual surveillance and monitoring systems in security-sensitive enviroments, gait recognition attracts more and more attentions as the modality can be captured at a distance.Gait recognition mainly analyzes moment images including walk style. The key factors are to find out the gait characters and classification method. It contains many kinds of techniques, such as computer vision, pattern recognition and so on. Focusing on this topic, this paper makes a research on gait feature extraction, recognition and evaluation. In this paper, we carry out the research work as follows:(1)This paper disscusses silhouette-based feature descriptor. Human silhouette geometry is generated by boundary tracking approach and resampled to a normalized format. Boundary-centroid distance is proposed to describe gait modality. Then, we apply wavelet transform to boundary-centroid distance, and extract wavelet descriptor. At the same time, we obtain the human skeleton model and extract body's dynamic parameters to express gait modality.(2) Gait cycle is estimated according to the variety signals of human width. We use the gait cycle as the basic unit of recognition and choose the maximum as the key point.(3) We carry out human identification based on SVM in different databases using the two kinds of gait feature. The performances based on the two features are compared. Multiple feature fusion and multiple views fusion are carried out and the recognition results demonstrate that the performance of multiple feature and multiple views recognition is better than any single feature and single view recognition.(4) Probability of Correct Recognition and Cumulative Match Scores are proposed to evaluate the performance of the recognition.Presently, research and applications about human recognition based on gait are still in the ascendant, and the achievement of this paper may play a certain role in the development of gait recognition.