Extraction and Temporal Segmentation of Multiple Motion Trajectories in Human Motion

We propose a new method for extraction and temporal segmentation of multiple motion trajectories in human motion. Motion trajectories are very compact and representative features for activity recognition. Our method extracts motion trajectories generated by body parts without any initialization or any assumption on color distribution. Tracking human body parts (hands and feet) are difficult because body parts which generate most of motion trajectories are relatively small in relation to the human body. We overcome this problem using our new motion segmentation method. We detect candidate motion locations in every frame and set these locations as Significant Motion Points (SMPs). We obtain motion trajectories by combining SMPs and the color-optical flow based tracker results. These motion trajectories are used as features for temporal segmentation of specific activities from continuous video sequences. The characteristics of motions trajectories as features make the separate activity recognition for different body parts possible. We tested our approach on actual ballet steps. Experimental results show that the proposed method can successfully extract and temporally segment multiple motion trajectories in human motion.