A new hand gesture recognition scheme for similarity measurement in a vision based barehanded approach

Vision based hand gesture recognition is an open research problem in Human Computer Interaction (HCI). The most common form of gestures are signs used in different sign languages having both local and global motions. Recognition of these gestures require the hand trajectory both with respect to orientation in space among inter frames and distance among objects like face and hands among intra frames. In this paper, at first the face region and the two hand regions are detected. The centre of the face along with the centre of the two hands are then determined. This is then followed by tracking the hands trajectory following the centre both in terms of direction among consecutive frames as well as distance from the centre of the frame. The variations in the hand locations along with the centre of the face are considered for analysis. Finally the gestures are considered for recognition. The paper addresses a pertinent problem w.r.t the difference in the number of frames between a stored gesture and query gesture for matching. Experimental results show an accuracy of 86% in recognizing 24 gestures is maintained even after normalizing the number of frames by the proposed scheme.