基于Kinect骨骼数据的手势识别

为提高手势识别系统的识别率和识别效率,提出基于Kinect骨骼数据的手势识别方法。在获取的关节坐标的基础上,统计各关节在空间3个维度上的变化量并表示为相应的权值,在动态时间规整(DTW)算法中进行特征加权,提高算法识别率和鲁棒性。在进行匹配前使用模板选择方法从数据库选取与待测手势最为相近的模板,减少用于DTW匹配的模板数量,缩短样本识别时间。实验结果表明,所提方法具有较高的识别率和识别效率。

Gesture recognition based on Kinect skeleton data

To improve the accuracy and efficiency of gesture recognition system,a gesture recognition method based on Kinect skeleton data was proposed.On the basis of the obtained joint coordinates,the variation of each joint in three dimensions of space was calculated and expressed as the corresponding weight,feature weighting method was performed in dynamic time warping (DTW) to improve the accuracy and robustness of algorithm.A template selection method was used to select templates from dataset,which reduced the number of templates used for the DTW and shortened the sample identification time.Experimental results show that the proposed method has high recognition accuracy and efficiency.