Head movement recognition based on LK algorithm and Gentleboost

摘要

Different head movements can help us not only understand accurately whether the speaker in question is in favor of or against the current topic but also recognize human expressions better, i.e. distinguishing similar expressions such as smile and scorn. Since the movements of the feature point — nostrils on the face and the movements of the head are almost the same, tracing the feature point with Lucas-Kanade(LK) algorithm we can find out the pattern of the head movements. We can train GentleBoost classifiers use the difference of coordinates of nostril in an image or a frame to identify the movement of head. A number of tests have been conducted and the results have proved the accuracy, reliability and efficiency of the method of discussed in the paper.