基于Kinect视觉功能对体感手势目标识别

手势识别技术在人机交互模式中是一项重要的研究课题,为提高手势识别性能,文章介绍一种基于Kinect体感设备对手势目标进行识别的方式。利用Kinect视觉传感器获取RGB彩色手势图像和深度图像,使用OpenNI从中识别人体关节并提取空间坐标,通过空间向量运算,完成手部定位并获取手部轮廓,结合动态规划思想的限幅滤波算法,得到最优解的手势识别。实验测试表明,该系统具有良好的实时性和鲁棒性,能够实现用手势动作控制PowePoint的自由播放。

Gesture recognition technology is an important research topic in human-computer interaction mode.In order to improve the performance of the gesture recognition,this paper describes a device based on Kinect somatosensory target gesture recognition mode.We used Kinect gestures vision sensor to get RGB color images and depth images;we used OpenNI to identify joints of the human body and extract its spatial coordinates;we adopted the space vector operations to complete hand positioning and get the hand contour;and through the clipping and filtering algorithm based on dynamic programming,we obtained the optimal solution to gesture recognition.The experimental tests show that the system has good real-time performance and robustness,thus realizing the free playing of PowePoint controlled with gestures.