基于自适应手指分割与判别的静态手势识别

STATIC GESTURE RECOGNITION BASED ON ADAPTIVE SEGMENTATION AND DISCRIMINATION OF FINGERS

由于动态手势可以看作是多帧静态手势的融合,研究静态手势成为解决手势识别问题的重点。针对静态手势,提出一种自适应手指分割与判别的手势识别算法。首先,运用YCb Cr颜色空间的肤色聚类特性对手势图像进行分割,并采用质心点漂移的理念确定手指方向并作旋转归一化处理;其次,针对手势轮廓点的梯度方向和跨度确定手指的候选区域,并结合形态学的方法重建出单一手指的二值化形态;最后,选取恰当的形状特征,运用SVM分类器对其形状特征进行分类。实验表明该方法具有较好的识别率。

Since dynamic hand gestures can be regarded as the fusion of multi-frame static gestures,thus static gesture study becomes the key of solving gesture recognition problem. Aiming at static hand gestures,this paper puts forward a gesture recognition method which is based on adaptive segmentation and discrimination of fingers. First,the method segments hand gesture image using skin colour clustering feature of YCb Cr colour space,and adopts the idea of centroid drift for finger direction determination and makes rotation normalisation processing. Secondly,it determines candidate area of fingers aimed at the gradient direction and span of gestures contour points,and restores in combination with morphological method the binary form of a single finger. Finally,it selects appropriate shape features and uses support vector machine( SVM) classifier to classify its shape features. Experimental results show that the method has promising recognition rate.