Hand tracking and fingertip detection based on Kinect

Abstract

With the rapid development of computer technology and sensor performance, hand gesture recognition (HGR) has become one of the most promising research in human computer interaction field. Among all the procedures of HGR, real-time hand tracking is of great importance, which also faces many existing challenges. In this paper, an effective method was proposed for hand tracking in real time using Kinect, including the hand region extraction and fingertips determination. During the former procedure, we presents a novel two-scan method based on connectivity analysis and recrusion algorithm to extract the hand region in real time. After this, in order to determine the fingertips, we first implemented a fast hand convex hull extraction algorithm to obtain the primary candidate points. Then two typical constraints based on hand shape characteristics are applied to determine the final fingertip points. The subsequent experiments verified that our proposed strategy could significantly improve the quality of real-time hand tracking.