Position estimation of leg inertial sensors to improve accuracy of stride length estimation

Abstract

Human gait analysis generally needs acceleration at the center of joints where inertial measurement units (IMU) cannot be attached. The joint acceleration estimation requires the position vectors from the IMU to the center of the joint in the sensor coordinate system. Past studies have proposed a method to estimate the position vector of the joint center from two IMUs attached to two rigid bodies, respectively. The idea is based on the equality of a joint center acceleration estimated from two IMUs. However, the estimation accuracy of the position vector depends on the calibration movement. The purpose of this study is to verify the accuracy of the position vector from a lower leg IMU to the ankle joint with calibration movements. We determined the movements as a squat movement and a unidirectional knee rotating movement and compared their accuracy. We assumed that the ankle joint keeps resting throughout the calibration movements. The result demonstrated that the accuracy was better for the unidirectional knee rotation movement and dependent on the quality of the calibration movement.