关键词: 深度信息, 指尖检测, 手势识别, 人机交互

Abstract: This paper presents an algorithm to realize the finger and palm tracking based on the depth information in real time. The recognized actions can be used for hand pose recognition. The depth map of the hands is captured by Kinect, which then converts the depth map to 3D point cloud in a form of pixel matrix. Fingers and direction of palms are calculated by using K-curvature algorithm. Finger names are determined according to their relative distances. Experimental results show that the present method effectively achieves the finger and palm tracking, and is unaffected by light and complex background. This method can apply to hand pose recognition.

Key words: depth data, fingertip detection, gesture recognition, Human-Machine Interaction（HCI）