关键词: 光流法, 步态能量图, 局部二值模式, 梯度方向直方图, 融合特征

Abstract: For the Gait Energy Image（GEI）, which only depicts the contour information while extracting the human body information, and loses the limitation of the internal features, a GEI recognition algorithm based on the Local Binary Pattern（LBP） and the Histograms of Oriented Gradients（HOG） is proposed. The algorithm steps include：firstly, the gait cycle is extracted by using the optical flow method, and a cycle of the Gait Energy Image（GEI） is obtained, then the LBP features of GEI are extracted in three layers, the LBP images of three layers are extracted, then the HOG features of each layer are abstracted, and then the LBP and HOG features are fused. The new features of each layer are obtained, and finally three new features are fused into the final features that can be used for recognition. Through experimental comparison of CASIA and USF gait database, this method has achieved higher recognition rate.

Key words: optical flow method, Gait Energy Image（GEI）, Local Binary Pattern（LBP）, Histograms of Oriented Gradients（HOG）, fusion features