关键词: 手势识别, 多特征, 肤色分割, 凸包算法, 决策树

Abstract:

Computer vision-based hand gesture recognition has become a key technique in the realization of new human-computer interaction.This paper proposes a method in the recognition of real-time hand gestures with multiple cues based on OpenCV.The skin area is detected with HSV,and the hand is segmented with gesture geometric characteristics.Then the fingertips are detected by the convex hull algorithm.For distinguishing and recognizing the 12 different gestures defined,a decision tree is adopted combined with the number of fingers,the angle between your fingers aspect ratio features and contour features.Experimental results demonstrate that the proposed method is robust and real-time,and high recognition rate and various gestures are achieved.

Key words: gesture recognition;multiple cues;skin segmentation;convex hull algorithm;decision tree