A Gesture Interaction System Based on Improved Finger Feature and WE-KNN
Feifei Li, Yujun Li, Baozhen Du, Hongji Xu, Hailiang Xiong, Min Chen · 2019
In most gesture recognition research fields, feature extraction is based on single finger. In this paper, we propose an improved finger feature extraction algorithm based on double fingers, and it is easier to judge the projected distance and angle of five fingertips and can be better applied to gesture recognition with multi-fingers' information. Furthermore, K-nearest neighbor (KNN) classifier based on entropy-weight allocation (WE-KNN) is proposed to improve the accuracy of gesture recognition. Compared with traditional algorithms, the proposed finger feature extraction algorithm combined with WE-KNN can enhance the accuracy of gesture recognition on Leap Motion dataset. What's more, we apply the proposed algorithms to the gesture interaction system of virtual gym scenario, which can increase the interactive fun and improve the interactive experience.