Human Behavior Recognition Based on Axonometric Projections and PHOG Feature

Human behavior recognition has become a hot research topic in computer vision. In this paper, we propose an effective method to recognize human behaviors from sequences of depth maps, which provide additional body shape and motion information for behavior recognition. In our approach, we construct a novel difference motion history image, and propose axonometric projection to capture the target motion process, after that, pyramid histogram of orientated gradients is extracted for each view to describe the target motion. Large scale experimental results on challenging and public MSR-action3D behavior dataset demonstrate that the performances of our difference motion history image and our descriptors are satisfying, what’s more, our proposed axonometric projection approach further improves the performance, which significantly outperforms the state-of-the-art methods.