利用Kinect手部跟踪构造手势空间点集进行指尖定位的研究与实现

Fingertip location through the gesture space points sets constructed by Kinect hand tracking

利用Kinect实现非接触式人机交互是当前研究的热点,手势识别是Kinect应用程序的核心,但是Microsoft Kinect for Windows SDK提供的手部深度图像跟踪算法中没有手指的描述子集,因此不能进行有效的指尖定位.基于此,本文提出利用Kinect手部跟踪点获取真实坐标,进而构造空间点集来提取手势信息,其中空间跟踪点集是指满足目标特征且与局部跟踪信息相关的空间坐标集合.在此基础上,利用K-曲率算法实现手势的指尖定位,并通过9个经典手势对该方法进行验证.实验结果表明,通过本文提出的实现方法所获取的指尖位置准确率可达到91%,指尖定位速度可以达到30FPS,完全可以满足体感游戏、页面自动换页等功能的需要.

Non-contact human-computer interaction was realized using Kinect.Gesture recognition is the core of Kinect applications.Effective finger positioning cannot be obtained because image tracking algorithm for hand depth provided by Microsoft Kinect for Windows SDK has no finger descriptor sets.This paper therefore proposed a method to obtain real coordinates by Kinect hand tracking point and by extracting gesture information constructing space point sets,the space tracking point sets is the spatial coordinates sets that meet the target characteristics and is associated with the local track information.Fingertip location is therefore implemented through K-curvature algorithm and verified by nine classic gestures.It was found that fingertip position accuracy reached 91% and the finger positioning speed reached 30 FPS proposed by the realization method in this paper,which could fully meet the needs of function such as somatosensory games,and automatic paging.