利用轨迹模板匹配方法的实时动态手势识别

摘要

利用混合高斯模型进行运动检测,分割出运动前景,采用粒子滤波器结合皮肤椭圆模型进行手势跟踪,获得手势中心点运动轨迹,在此基础上提出利用轨迹模板匹配方法进行动态手势识别.该方法利用基本的几何和三角函数就能完成手势运动轨迹的定义和识别,不需要选择特征或训练样本.实验结果表明,该算法能够实现实时动态手势识别.%The Gaussian mixture model is used to segment the motion foreground.A particle filter anda skin ellipse model are used to track the motion of the center of the gesture.Based on this,the dynamic gesture recognition is proposed using the track template matching method.This method can use the basic geometric and trigonometric functions to complete the recognition of the trajectory of the gesture,without selecting features or training samples.The experiment results show that the algorithm can achieve real-time dynamic gesture recognition.