当前位置: X-MOL 学术Inform. Fusion › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A sensor fusion approach to MARG module orientation estimation for a real-time hand tracking application
Information Fusion ( IF 18.6 ) Pub Date : 2022-09-24 , DOI: 10.1016/j.inffus.2022.09.017
Neeranut Ratchatanantakit, Nonnarit O-larnnithipong, Pontakorn Sonchan, Malek Adjouadi, Armando Barreto

This paper introduces a new algorithm (Gravity & Magnetic North Vector correction — Double SLERP, or “GMV-D”) to estimate the orientation of a MEMS Magnetic/Angular-Rate/Gravity (MARG) sensor module using sensor fusion in the context of a real-time hand tracking application, for human–computer interaction purposes.

Integrated MEMS MARG modules are affordable, small, light and consume minimal power. As such, there is interest in using them for monitoring the orientation of various body segments, to which they can be attached (e.g., the finger segments of a gloved hand). However, each of the 3 types of signals they provide has proven insufficient to yield robust orientation estimates, particularly in regions of space where the geomagnetic field is distorted. The significance (main contribution) of the approach we present is the computation of a final orientation estimate that uses all the signals generated by inexpensive (e.g., less than 20 USD, in large quantities) integrated MEMS MARG modules but weighs their contributions with simple, real-time-updatable parameters that prevent erroneous corrections when the pre-conditions for their valid use are not met. This will enable the use of inexpensive integrated MEMS MARG modules for hand tracking applications in human–computer interaction and other areas of work where tracking the orientation of body segments in real-time is important.

In each iteration, GMV-D defines an initial orientation estimate from integration of gyroscopic signals (“dead reckoning”), and also calculates accelerometer-based and magnetometer-based corrections. These corrections are defined on the assumptions that the module is near static and affected by an undistorted geomagnetic field.

Because these assumptions are seldom fully met simultaneously, the information fusion challenge is to apply each correction only to the extent that its corresponding pre-conditions are met, as inappropriate corrections will introduce significant error in the future orientation estimates. To achieve this, GMV-D develops an accelerometer correction trustworthiness parameter, 0 α 1, and a magnetometer correction trustworthiness parameter, 0 μ 1, both of which are updated on a sample-by-sample basis and are available at each iteration of the algorithm. The information fusion phase of the algorithm implements the corrections in a two-tiered application of Spherical Linear Interpolation (SLERP) of the quaternions representing the initial dead reckoning estimate and the available corrections, scaled according to their corresponding levels of trustworthiness.

GMV-D was evaluated in comparison to 2 other orientation correction approaches (Kalman Filtering and GMV-S) and contrasted with 2 contemporary complementary filter approaches (Madgwick , Mahony). The results confirm that GMV-D displayed better orientation estimation performance when the algorithms operated in an area with known distortion of the geomagnetic field.



中文翻译:

用于实时手部跟踪应用的 MARG 模块方向估计的传感器融合方法

本文介绍了一种新算法(重力和磁北矢量校正 — 双 SLERP,或“GMV-D”),在用于人机交互目的的实时手部跟踪应用程序。

集成 MEMS MARG 模块价格实惠、体积小、重量轻且功耗最低。因此,使用它们来监测它们可以附着到的各种身体部分(例如,戴手套的手的手指部分)的方向是有意义的。然而,事实证明,它们提供的 3 种类型的信号中的每一种都不足以产生可靠的方向估计,特别是在地磁场失真的空间区域中。我们提出的方法的意义(主要贡献)是计算最终方向估计,该估计使用由廉价(例如,少于 20 美元,大量)集成 MEMS MARG 模块生成的所有信号,但用简单的方法衡量它们的贡献,可实时更新的参数,可在不满足有效使用的先决条件时防止错误更正。

在每次迭代中,GMV-D 根据陀螺仪信号的积分(“航位推算”)定义初始方向估计,并计算基于加速度计和基于磁力计的校正。这些修正是在假设模块接近静态并受未失真地磁场影响的假设下定义的。

由于这些假设很少同时完全满足,信息融合的挑战是仅在满足其相应先决条件的情况下应用每个校正,因为不适当的校正将在未来的方向估计中引入重大错误。为此,GMV-D 开发了加速度计校正可信度参数,0α1、磁力计校正可信度参数,0μ1,两者都在逐个样本的基础上更新,并且在算法的每次迭代中都可用。该算法的信息融合阶段在表示初始航位推算估计的四元数和可用校正的球面线性插值 (SLERP) 的两层应用中实现校正,并根据其相应的可信度级别进行缩放。

GMV-D 与 2 种其他方向校正方法(卡尔曼滤波和 GMV-S)进行了比较评估,并与 2 种当代互补滤波器方法(Madgwick,Mahony)进行了对比。结果证实,当算法在已知地磁场失真的区域中运行时,GMV-D 显示出更好的方向估计性能。

更新日期:2022-09-24
down
wechat
bug