Efficient and effective human action recognition in video through motion boundary description with a compact set of trajectories

Human action recognition (HAR) is at the core of human-computer interaction and video scene understanding. However, achieving effective HAR in an unconstrained environment is still a challenging task. To that end, trajectory-based video representations are currently widely used. Despite the promising levels of effectiveness achieved by these approaches, problems regarding computational complexity and the presence of redundant trajectories still need to be addressed in a satisfactory way. In this paper, we propose a method for trajectory rejection, reducing the number of redundant trajectories without degrading the effectiveness of HAR. Furthermore, to realize efficient optical flow estimation prior to trajectory extraction, we integrate a method for dynamic frame skipping. Experiments with four publicly available human action datasets show that the proposed approach outperforms state-of-the-art HAR approaches in terms of effectiveness, while simultaneously mitigating the computational complexity.