Natural human-computer interaction using static and dynamic hand gestures

The paper presents design and implementation issues of a natural gesture user interface for real-time hand gesture recognition in depth data collected by a Kinect sensor. The hands are first segmented based on depth and a novel algorithm improves the scanning time in order to identify the first pixel on the hand contour within this space. Starting from this pixel, a directional search algorithm allows for the identification of the entire hand contour and the k-curvature algorithm is then employed to locate the fingertips over it. Dynamic time warping (DTW) guides the selection of gesture candidates and also recognizes gestures by comparing an observed gesture with a series of pre-recorded reference gestures. The comparison of results with state-of-the-art approaches shows that the proposed system is similar in performance or outperforms the existing solutions for the static and dynamic recognition of signs.