Temporal Human Body Motion Segmentation

T. Jin, M.K.H. Leung, and L. Li (Singapore)


Temporal difference, rear motion history image, front motion history image, motion vector.


Human motion analysis is receiving great interests from computer vision researchers. To recognize human behaviors, pixel intensities are transformed into features to represent different actions. Hence, the analysis of feature pattern (i.e. motion pattern) plays a vital role for human action recognition. In this paper, a novel approach to generate motion-based pattern is proposed. Moving objects are firstly segmented by employing the adaptive threshold based change detection approach. Secondly, we use scalar valued rear motion history image (RMHI) and front motion history image (FMHI) to represent how motion is evolved. This representation is then used to segment and measure the motion induced by the object. Thirdly, the motion vectors with orientation and magnitude are generated from the chamfer distance. Finally, we derive one novel approach to generate intra motion history image (IMHI) for inside moving parts. The experimental results are encouraging.

Important Links:

Go Back