Quaternion-based pseudo kalman filter for wearable inertial/magnetic sensor applications

Jung Keun Lee

doi:10.1051/matecconf/20165901009

Abstract

This paper deals with orientation estimation using miniature inertial/magnetic sensor comprised of a tri-axial rate gyro, a tri-axial accelerometer, and a tri-axial magnetometer. Particularly, a novel quaternion-based pseudo Kalman filter (KF) is proposed by modifying an indirect KF, in order to maximize the computational efficiency and implementation simplicity. In the proposed pseudo KF, time-update process for prediction is based on the quaternion itself, while measurement-update process for correction is performed through the quaternion error. Experimental tests were conducted to verify performance of the proposed algorithm in various dynamic conditions. By designing the pseudo KF structure, matrix operations required in a typical KF are simplified. For instance, the proposed KF does not require the evaluation of the a priori and a posteriori error covariance matrices. Thus, the proposed algorithm achieves higher computational efficiency even than a typical indirect KF, without sacrificing estimation accuracy. Due to its high efficiency, the proposed algorithm can be suitable for battery-powered and low cost processor-based wearable inertial/magnetic sensor applications.

Highlights

With the advances in micro-electro-mechanical systems (MEMS) technology, miniature inertial/magnetic sensors have emerged for wearable motion capture technology [1,2]
This paper proposes a quaternion-based pseudo Kalman filter by modifying an indirect KF, in order to maximize the computational efficiency
It is important to note that the proposed pseudo Kalman filter is realized only because the Kalman state is 'small' quaternion error

Summary

Introduction

With the advances in micro-electro-mechanical systems (MEMS) technology, miniature inertial/magnetic sensors have emerged for wearable motion capture technology [1,2]. An inertial/magnetic sensor is typically comprised of a tri-axial rate gyro, a tri-axial accelerometer, and a tri-axial magnetometer, and is used to estimate accurate 3D orientation in an unconstrained environment (e.g., outdoors), associated with physical medicine and rehabilitation applications [3,4]. In terms of the orientation representation, Euler angles [5], quaternion [1,2,6,7], and direction cosine matrix [8] are common. The quaternion can arguably be considered as the most popular representation for 3D orientation estimation due to its singularity-free aspect and relatively small number of dimensions (i.e., four). One approach is the direct KF which designates the quaternion itself as the state. The other approach is an indirect KF which does the quaternion error as the state [9]

Methods

Results

Conclusion