Depth camera based real-time fingertip detection using multi-view projection

Abstract

We propose a real-time fingertip detection algorithm based on depth information. It can robustly detect single fingertip regardless of the position and direction of the hand. With the depth information of front view, depth map of top view and side view is generated. Due to the difference between finger thickness and fist thickness, we use thickness histogram to segment the finger from the fist. Among finger points, the farthest point from palm center is the detected fingertip. We collected over 3,000 frames writing-in-the-air sequences to test our algorithm. From our experiments, the proposed algorithm can detect the fingertip with robustness and accuracy. © 2013 Springer-Verlag.