基于Kinect的手势数据库及动静态手势识别算法

Hand gesture based computer human interaction has received continuous focuses by researchers. The increasing interests on this field demand a comprehensive hand gesture data set. We introduce a Kinect-based hand gesture data set and a rule-based static and dynamic hand gesture recognition algorithm. The hand gesture data set, named as SCUT-DCHG, includes depth and color sequences of nine static and thirteen dynamic hand gestures performed by multiple persons standing in various distances from Kinect. Moreover, we propose a new real-time rule-based dynamic and static hand gesture recognition algorithm by extracting distinguishable features and classifying them according to various rules. Our experimental results reached 95.5% recognition accuracy for all gestures, specifically 96.2% and 94.4% for dynamic and static hand gesture individually, using SCUT-DCHG data set. The effectiveness of the proposed algorithm is shown.