Similar hand gesture recognition by automatically extracting distinctive features

Abstract

With the flourish development of computer vision technology, hand gesture recognition plays a more and more vital role in human-computer interaction for its convenient and nonverbal communication. However, confusion caused by similar gestures brings inherent errors when considering enough meaningful gestures in the database. In this paper, an automatic feature extraction for similar gesture recognition is proposed with respect to confusion arising in similar gestures. Except the orientation feature, four additional innovative features are extracted to distinguish all the similar gestures remarkably in the experimental database containing 10 numbers and 26 letters. Compared with the conventional method that a couple of similar gestures are extracted as a specific feature, the proposed method distinguishes similar gestures with automatic distinctive feature extraction. Experimental results show high recognition rate and versatility of the proposed method.