Abstract: K-Curvature and convex hull theory based finger detecting algorithm always suffering inaccurate fingertip capture. A new fingertip and finger pointing detection method was proposed which was based on finger contour features. There were three key steps in the method: contour smoothing， linear region detection， and fingers pointing and tips detection. Accurate and fast fingertip recognition processing could be realized. The proposed algorithm runs fast and accurately in fingertip recognition process. Through the experimental verification of different gestures， the fingertips position could be accurately located and the fingers pointing information could be obtained by the proposed algorithm. It can also meet the real-time requirements in this type of human-computer interaction system， and has strong adaptability to different finger direction.

Key words: fingertip detection, finger direction, gesture recognition, human-computer interaction, contour smoothing