基于Kinect的动态手势识别

Dynamic hand gesture recognition based on Kinect

针对现有的动态手势识别率低,识别手势少等不足,利用Kinect设备提出了动态手势识别算法。首先利用Kinect捕获人的手部区域,采用基于像素分类的指尖检测算法找到指尖的个数,并以左右手的手指个数作为动态手势的开始和结束;对人手的运动轨迹进行分析,针对运动轨迹的运动方向的变化,提取了该动态手势的运动方向变化角度作为特征;采用隐马尔科夫模型训练和识别各个手势。实验结果表明:方法能够识别16个大写手写英文字母,且效果较好。

In view of shortcomings of the existing dynamic hand gesture recognition,a dynamic hand gesture recognition algorithm based on Kinect device is proposed. The hand area is got through Kinect and detect the number of fingers,numbers of the left and right hand fingers is used as start and end of dynamic hand gesture.Motion trajectory of the hand is analyzed,aiming at movement direction change of movement track,and extract change-of-direction angles of dynamic gesture as feature. Use hidden Markov model（ HMM） train and recognize each gesture. Experimental result proves that using this method can recognize 16 capital hand-writen English letters,and the effect is good.