一种基于改进DTW算法的动态手势识别方法

结合Kinect传感器提出了一种基于改进的DTW算法的动态手势识别方法。首先，通过SDK对Kinect传感器获得的深度信息进行分析，获得人体骨骼点3D位置，选取其中4个点作为手部运动特征；然后，用加权距离和全局路径限制的方法对传统的DTW算法进行改进；最后，用改进的DTW算法进行模板训练和实现动态手势的识别。实验表明：该方法能很好地实现动态手势的识别，实时性好，在背景干扰和光照方面有较强的鲁棒性，较传统的DTW算法在识别速度和识别正确率方面有所提高。

Novel dynamic gesture recognition method based on improved DTW algorithm

A dynamic gesture recognition method is proposed by combining the Kinect sensor in this paper. This method is based on the improved DTW. Firstly, the depth information is analyzed through the SDK. The depth information is obtained by Kinect sensor. The human skeletal point 3D position is obtained, and the four points are selected as the characterized hand movement. Then the traditional DTW algorithm is improved by weighted distance and restricted global path. Finally, the improved DTW algorithm is used in the template training and dynamic gesture recognition. The results indicate that this algorithm can achieve dynamic gesture recognition and is real-time. In the background interference, the method is robust in the light. Compared with the traditional DTW algorithm, the recognition speed and recognition accuracy are improved.