基于光流的动态手势识别

Dynamic Hand Gesture Recognition Based on Optical Flow

手势识别是人机交互中的重要组成部分,文章针对基于光流PCA(主分量分析)和DTW(动态时间规整)进行命令手势识别。利用块相关算法计算光流,并通过主分量分析得到降维的投影系数,以及手掌区域的质心作为混合特征向量。针对该混合特征向量定义了新的加权距离测度,并用DTW对手势进行匹配。针对9个手势训练和识别,识别率达到92%。

Hand gesture recognition is an important component of human-computer interaction. This paper recognizes command gesture-based optical flow PCA (principle component analysis) and DTW (dynamic time warping). Optical flow is computed with block relative algorithm, and then processed with PCA to form low-dimensional projection coefficients. The coefficients are combined with the center position of hand as mixed feature vector. A new weighted distance is defined for this mixed vector, which is then used in DTW for hand gesture matching. The recognition rate on 9 gestures is 92%.