Gabor-Based Kernel PCA with Fractional Power Polynomial Models for Gait Recognition

Abstract

In this thesis we propose a method to extract the human gait features from the surveillance video through Gabor wavelet transformation and then we classify these features by kernel principle component analysis (PCA) with the fractional power polynomial model Because human gait feature extraction can be categorized into spatial and temporal domain we will discuss the gait features in these two domains In order not to lose any information from the surveillance video this thesis uses the spatial-temporal silhouette of the people walking in the surveillance video then we can have the gait features by taking silhouette convolution with Gabor based wavelet transformation We classify these features by kernel PCA with the fractional power polynomial model Finally we use Mahalanobis distance to measure the similarity between the gait features The simulation and the experiment results show that Gabor-based kernel PCA with fractional power polynomial models for Gait recognition have a better performance

Date of Award	2015 Aug 6
Original language	English
Supervisor	Tzuu-Hseng S. Li (Supervisor)