Statistical registration and modeling of frontal view gait data with application to the human recognition

摘要

We study the problem of analyzing and classifying frontal view human gait data by registration and modeling on a video data. We suppose that frontal view gait data as a mixing of scale changing, human movements and speed changing parameter. Our gait model is based on human gait structure and temporal-spatial relations between camera and subject. We estimate the parameters of the human gait using the multistep algorithms based on the method of nonlinear least squares. Proposed algorithm is very stable to estimate each parameters. Finally, we apply a k-nearest-neighbor classifier, using the estimated parameters, to perform human recognition, and present results from an experiment involving 120 subjects. As a result, our method shows high recognition rate, that has the better performance compared to other methods.