Speed invariance vs. stability: cross-speed gait recognition using single-support gait energy image

Chi Xu, Yasushi Makihara, Xiang Li, Yasushi Yagi, Jianfeng Lu

Abstract: Gait recognition has recently attracted much attention since it can identify person at a distance without subject cooperation. Walking speed changes, however, cause gait changes in appearance, which significantly drops performance of gait recognition. Considering a speed-invariant property at single-support phases where stride change due to speed changes are mitigated, and a stability against phase estimation error and segmentation noise by aggregating multiple phases inspired by gait energy image (GEI), we propose a speed-invariant gait representation called single-support GEI (SSGEI), which realizes a good trade-off between the speed invariance and the stability by combining single-support phases and GEI concept. For this purpose, we firstly find out the optimal duration around single support phases using a training set so as to well balance the speed invariance and the stability. We then extract SSGEI by aggregating multiple single-support frames. Finally, we combine the proposed SSGEI with subsequent Gabor filters and metric learning for better performance. Experiments on the publicly available OU-ISIR Treadmill Dataset A composed of the largest speed variations demonstrated that the proposed method yielded 99.33% rank-1 identification rate on average for cross-speed gait recognition, which outperforms the other state-of-the-arts, and realized a low computational cost as well.