基于VR设备中IMU的头部姿态感知算法研究

Head Attitude Sensing Algorithm based on an IMU in VR Device

摘要:

随着头戴式显示设备的发展,基于虚拟现实(Virtual Reality,VR)的教育培训随之流行开来。在基于VR设备的教育培训中,存在用户与设备间进行交互的场景。在这些场景中,VR设备对使用者头部姿态的感知尤为重要。为了保证较高的姿态解算精度,同时降低系统的计算量,设计了四种基于头戴式惯性测量单元(Inertial Measurement Unit,IMU)的姿态解算方法,并对比了这四种算法的姿态解算精度和运算效率。实验结果表明,相比于其他三种算法,使用四元数微分方程的三阶泰勒展开递推式更新四元数,同时利用间接扩展卡尔曼滤波器融合地磁信息进行修正的姿态解算方法保证了较高的解算精度和较少的运算时间。该基于地磁修正+三阶泰勒展开法的头部姿态感知算法计算所得欧拉角与SBG公司生产的IG-500N型号IMU中提供的姿态角具有1.1×10-2度的总体平均偏差，且该算法使用MATLAB R2013a平台计算14000组数据的耗时为5.1秒。

Abstract:

With the development of head-mounted display devices, education and training based on virtual reality (VR) become popular. For education and training based on VR devices, there are interaction behaviors between users and devices. In these scenarios, it is particularly important for the VR devices to sensing the attitude of users. In order to ensure high-accuracy attitude resolution and reduce the systematic calculating cost, four attitude resolution methods based on the head-mounted inertial measurement unit (IMU) are designed in this paper. The comparison experiments of the four algorithms are given to evaluate their accuracy and efficiency. The experimental results demonstrate that the algorithm using third-order Taylor expansion recursive method to calculate the quaternion and the indirect extended Kalman filter to fuse the geomagnetic information outperforms the other three algorithms for the highest accuracy and the least operation time. The Euler angle calculated by the head attitude sensing algorithm based on the geomagnetic correction + third-order Taylor expansion method has an overall average deviation of 1.1 × 10-2 degrees from the attitude angle provided by the IG-500N IMU produced by SBG corporation, and the the time consumption of the algorithm is 5.1 seconds when 14000 sets of data are computed using MATLAB R2013a platform.