IMU-Based Knee Angle Estimation using an Extended Kalman Filter

Abstract

Several robotic-based rehabilitation therapies for hemiparetic patients need from monitoring the healthy leg using ambulatory systems such as Inertial Measurement Units (IMUs). However the use of these sensors involves several drawbacks like numerical drift, instable measurements in presence of magnetic fields or crucial alignment of the sensors. Based on a recursive paradigm to estimate the rotational axis, we present in this paper an algorithm to estimate the knee angle based on geometrical constrains using an Extended Kalman Filter. We tested this algorithm in five heathy subjects and were able to reconstruct the waveform of the knee angle with an average rms error lower than 2°.