改进型手姿态关键点估计方法在Leap Motion中的应用

Application of Improved Hand Gesture Key Point Estimation Method in Leap Motion

在虚拟现实和人机交互交叉领域中,人手姿态的精确估计显得尤为重要。通过对基于Leap Motion的手姿态关键点模型的研究,提出一种改进型的手姿态关键点估计方法。该方法基于三维骨架模型所提出的解三角形或解四边形的方法,通过采集大量手姿态作为样本,并进行特征提取,得到各指间关键点的位置及角度关系作为参数,将关键点连成一个网状曲面,从而形成一种新的通过指间关系进行关键点估计方法。实验表明,改进后的方法在关键点的位置估计准确率达到98%,明显提高了手姿态关键点估计得精确性和鲁棒性,并且通过各指间关键点的变化作为手姿态识别的预判,为动态手势识别方法的改进奠定了基础。

This paper proposes an improved key point estimation method for hand gesture.The method is based on three-dimensional matrix model of the proposed solution triangular or quadrilateral method,by collecting a large number of hand gestures as a sample,and feature extraction to obtain the position and angle relationship between the fingers of key points as parameters,put the key points together into a mesh surface,forming a new relationship between the fingers through the key points estimation.