Extracting key postures in a human action video sequence | IEEE Conference Publication | IEEE Xplore