Kinect骨骼信息下的动态手势识别研究

Dynamic Gesture Recognition Research Using Kinect's Skeleton Information

针对复杂环境中存在的手势识别问题,提出一种利用Kinect传感器获取深度信息并进行动态手势识别的方法。该方法通过对Kinect传感器获取的深度信息进行分析,获取人体主要骨骼点的3D坐标,从中选取六个点作为手部运动的特征参照;为提高手势识别系统的识别速度,提出了一种基于查表的DTW算法对得到的特征数据进行模板训练并实现动态手势识别。实验结果表明:该方法具有较高的识别速度和识别率,对复杂背景及光照强度变化具有较强的鲁棒性。

For gesture recognition problems that exist in complex environments,a dynamic gesture recognition method based on depth information using Kinect sensor was proposed. In the method,the depth information was analyzed that acquired from the Kinect sensor,and the 3D coordinates of the major human skeletal point can be gotten. Six points as the reference of hand movement characteristics were selected; In order to improve the rate of recognition and the identification speed of the system,a DTW algorithm look-up table was used in the template training and dynamic hand gesture was recognized. The experimental result show that： the method has a high identify speed and recognition rate for the dynamic gesture. At the same time,it also has strong robustness for the complex background and light intensity change.