基于Kinect深度信息的手势提取与识别研究

针对基于视觉的手势识别技术对环境背景要求较高的问题,提出了一种使用深度信息进行手势提取和识别的研究方案。采用了微软Kinect摄像头进行手势深度图的采集,再将深度图转换为三维点云,根据深度信息过滤来提取手势数据。对手势数据进行方向校正后统计手势数据中深度信息的区间分布特征并输入到支持向量机进行训练,从而实现了对数字手势1～5的手势识别。实验结果证明,该手势识别方案的平均识别率达到95%,使用设备简单且精度较高,鲁棒性较好。

Gesture extraction and recognition research based on Kinect depth data

Aiming at the problem that gesture recognition technology based on vision required a lot on environment and background, this paper presented a gesture extraction and recognition scheme based on depth data. It utilized Microsoft Kinect to capture gesture depth map, converted the depth map to 3D point cloud, and then used depth map filter to obtain gesture data. After direction adjustment for the gesture, calculated and imported the interval distribution feature of gesture depth information to support vector machine for training, thus implementing gesture recognition for number gesture 1 to 5. Experimental results show that the average recognition rate of this scheme is 95%, and the scheme makes high precision with simple device and good robustness.