Abstract
In view of the complexity of gait recognition and the unsatisfactory recognition effect in different states, a new method of gait recognition based on dynamic and static features fusion is proposed. In order to retain as many gait characteristics as possible, the gait energy graph is chosen as the static feature, stride and step frequency as dynamic features, and feature fusion is carried out. A method of kernel principal component analysis combined with local preserving projection to reduce dimension of gait feature (KPCA-LPP) is proposed. Aiming at the problem of low recognition rate in non-regular perspective and multiple-view recognition, the paper improves the multi-view gait network, and uses the fast decomposition orthogonal matching tracking algorithm based on kernel function to classify and identify. The simulation results show that the proposed algorithm has some advantages over the existing classical algorithms, good recognition rate, the recognition effect is not affected by the interference factors such as backpack and overcoat, and the robustness is better.