Motion and color cues for hands detection in video based gesture recognition

The paper addresses the problem of hands detection in video sequences as the first step for gesture recognition based applications. Gesture recognition emerges as a natural human computer interface technology and reliable hands detection algorithms are needed as preliminary processing steps. Discussions are conducted on several techniques including motion and color as cues for detecting hands in real time video. Background subtraction algorithms and skin color detectors are investigated. An approach of complexity O(n*log(n)) that interpolates standalone results using each cue is equally proposed, where n is the dimension of a given video frame.