Parsing 3D motion trajectory for gesture recognition

Motion trajectories have been widely used for gesture recognition. An effective representation of 3D motion trajectory is important for capturing and recognizing complex motion patterns. In this paper, we propose a view invariant hierarchical parsing method for free form 3D motion trajectory representation. The raw motion trajectory is first parsed into four types of trajectory primitives based on their 3D shapes. These primitives are further segmented into sub-primitives by the proposed shape descriptors. Based on the clustered sub-primitives, trajectory recognition is achieved by using Hidden Markov Model. The proposed parsing approach is view-invariant in 3D space and is robust to variations of scale, temporary speed and partial occlusion. It well represents long motion trajectories can also support online gesture recognition. The proposed approach is evaluated on multiple benchmark datasets. The competitive experimental results and comparisons with the state-of-the-art methods verify the effectiveness of our approach.