基于深度信息的手势识别算法研究

针对在复杂背景中传统手势识别算法的识别率低问题,利用Kinect的深度摄像头获取深度图像,分割出手势区域后进行预处理;提取手势的几何特征,并提出深度信息的同心圆分布直方图特征,融合手势的几何特征和深度信息的同心圆分布直方图特征;学习训练随机森林分类器进行手势识别.文中通过在复杂背景条件下对常见的“石头”、“剪刀”、“布”3种手势进行测试,实验结果表明:文中所提方法具有很好的平移,旋转和缩放不变性,能适应复杂环境的变化.

Research on gesture recognition algorithm based on depth information

In order to deal with the problem of low recognition rate of traditional hand gesture recognition algorithm with complex background,a gesture recognition algorithm based on depth information is proposed.Depth image is obtained by the depth camera of Kinect,pre-processing is performed after dividing the gesture area.Geometric features of hand gesture are extracted,character of concentric distribution histogram of depth information is presented.Learn and train random forest classifier for gesture recognition,the geometric features of the hand gestures and the histograms of the concentric circles of the depth information are used to recognize the tested hand gesture via random forest classifier.Three kinds of gestures,such as stone,scissors and cloth,are tested under complex background.The experimental results show that the proposed method has better translation,rotation and scaling invariance,and can adapt to the change of complex environment.