Statistical Evaluation of Orientation Correction Algorithms in a Real-Time Hand Tracking Application for Computer Interaction
Statistical Evaluation of Orientation Correction Algorithms in a Real-Time Hand Tracking Application for Computer Interaction
复制标题
计算机交互实时手部跟踪应用中方向校正算法的统计评估
DOI：
10.1007/978-3-031-05409-9_8
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
N. Ratchatanantakit, N. O-larnnithipong
中科院分区：
文献类型：
--
作者：
N. Ratchatanantakit, N. O-larnnithipong
在手部运动跟踪系统中，对微型磁角速率重力（MARG）模块的方向估计进行了统计评估。30名人类受试者进行了一项实验，以验证所提出的方向校正算法在非磁失真（MN）和磁失真（MD）区域的性能。Kruskal-Wallis检验表明，在非磁畸变区域，重磁双SLERP（GMV-D）定向校正算法、重磁单SLERP（GMV-S）定向校正算法和卡尔曼滤波器（KF）定向校正算法的效果相似。然而，统计测试表明，当在磁失真区域中操作时，由三种方法产生的取向估计的误差水平显著不同，所提出的GMV-D方法在三个欧拉角Φ、θ和ψ中产生较低水平的误差。这表明GMV-D方法能够更好地提供方位估计，该方位估计对可能存在于MARG模块的操作空间中的磁场的局部干扰更鲁棒。
A new approach to correct the orientation estimate for a miniature Magnetic-Angular Rate-Gravity (MARG) module is statistically evaluated in a hand motion tracking system. Thirty human subjects performed an experiment to validate the performance of the proposed orientation correction algorithm in both non-magnetically distorted (MN) and magnetically distorted (MD) areas. The Kruskal-Wallis tests show that the orientation correction algorithm using Gravity and Magnetic Vectors with Double SLERP (GMV-D), the correction using Gravity and Magnetic Vectors with Single SLERP (GMV-S) and the on-board Kalman-Filter (KF) performed similarly in non-magnetically distorted areas. However, the statistical tests show that, when operating in the magnetically distorted region, the level of error in the orientation estimates produced by the three methods is significantly different, with the proposed GMV-D method yielding lower levels of error in the three Euler Angles Phi, Theta and Psi. This indicates that the GMV-D method was better able to provide orientation estimates that are more robust against local disturbances of the magnetic field that might exist in the operating space of the MARG module.