基于方向直方图的手语识别方法研究

Recognition of Chinese Sign Language Based on Orientation Histogram

传统的手语识别方法基本都是利用离散的各帧静态图像进行识别,存在一定局限性,根据普通摄像头获得的视频图像,并采用方向直方图来获得单帧的静态特征矢量和各帧图像间的动态特征矢量,实现手语的识别。首先针对头两帧图像,通过手部边缘轮廓提取算法找到手的区域,然后从中提取出能表现手部形状的静态特征矢量。同时,对连续帧的图像做动作评估,获得手部移动的动态特征矢量。最后,将手部形状的静态特征与动态特征结合,采用使用欧氏距离作为矢量间匹配程度的度量算法以实现手语识别。实验对5个人的5种手语分别进行测试,均能正确识别,结果验证了该方法的有效性。

Traditional recognition of Chinese sign language mostly uses static feature vectors. This paper proposes an algorithm that extracts efficient feature vectors to recognize a hand gesture for sign language. The proposed algorithm recognizes hand gesture based on visual information without using any special gesture glove. To recognize hand gesture, the proposed method is divided into three steps. First, by using an edge - based hand area search algorithm, a hand block is found and segmented efficiently from the monochrome input images. Secondly, if hand area is successfully extracted, the feature vectors representing the hand shape are analyzed applying orientation histogram scheme. Also, the feature vectors of moving hand is obtained by motion estimation. Lastly, the hand gesture is recog- nized by feature vectors of hand＇ s shape and movements. The proposed algorithm can not only segment the hand area, but also extract the feature vectors from the gray scaled motion images representing language words.