一种联合几何特征和隐马尔可夫模型的手势识别算法

Gesture Recognition Algorithm Combined with Geometric Features and HMM

为了实现复杂背景下的手势识别,并提高手势识别的精度和效率,提出了一种联合几何特征和隐马尔可夫模型(HMM)的手势识别算法。文中首先建立YCbCr颜色空间,使用肤色检测与区域标识法进行自适应手势分割;然后使用凸包算法检测指尖点和指蹼点,提取手部空间分布几何特征;最后建立隐马尔可夫模型,通过阈值判定完成手势识别。实验测试结果表明,在保证不同手势的高检测率和低误检率的同时,文中算法的旋转、平移和缩放不变性也得到了进一步提高,具有高鲁棒性与低复杂度。

In order to achieve gesture recognition in complex background,and improve the accuracy and efficiency of gesture recognition,a gesture recognition algorithm combined with geometric features and HMM is proposed.First,the YCbCr color space is established using the method of skin color detection and component labeling to realize adaptive gesture segmentation.Then the fingertips and concave points are detected based on convex hull algorithm to obtain distribution features of a hand.Finally,the HMM model is established,and multidimensional statistics of distribution characteristics are used to complete the gesture recognition.Experimental test results indicate that the high detection rate and low false detection rate of different gestures not only are realized but also the rotation,translation and scale invariance are further improved,it verifies the high robustness and low complexity of the proposed algorithm.