Hand gesture interface based on improved adaptive hand area detection and contour signature

Abstract

HMD (head-mounted display) as a promising device is becoming more and more important in daily life. Many companies has been working on it for the next generation human-interface system. This paper presents a real-time hand gesture interface based on TSL (Hue, Saturation, Luminance) adaptive area detection and distance signature with single camera. First, apply self-adaptive skin color detection in TSL color space where skin color data can be clustered to segment hand area. Second, acquire the distance signatures from hand shape contours and obtain possible finger points which reduce the hand gesture recognition problem into finding peaks of one dimensional signature. Last, finger points are labeled by the information of signature. ROC (Receiver Operating Characteristic) Analysis shows the proposed hand area detection method always gives a result in feasible area (TPR>0.91, FPR<;0.1) which is suitable for the following contour analysis, indicating that it's more stable and robust compared with other skin color based methods. The evaluation results show the potential of real-time on PC at around 10 fps.