Gait Energy Image Based on Static Region Alignment for Pedestrian Gait Recognition

The Gait Energy Image (GEI) spatially aligns, accumulates, and averages all the frames of a gait cycle, so there is a very high requirement for the registration of moving targets. Accurate registration of moving targets is important for the synthesis of Gait Energy Image (GEI). In this paper, we propose a new Gait Energy Image to improve the registration effect: Gait Energy Image based on static region alignment (SRA-GEI). Firstly, we select the minimum circumscribed rectangle containing the moving human body from the gait sequence. Secondly, we scale the minimum circumscribed rectangle to the specified height and calculate the gait cycle by analyzing the distance between the two feet. Finally, we propose a new registration method to generate Gait Energy Image by calculating and aligning the centroid of the static region of the gait image. This paper explores the performance of SRA-GEI with KNN based on the CASIA Dataset B. The experimental results have shown that the proposed method achieves better recognition rate compared with GEI which aligned by overall centroid.