Position Estimation of an IMU Placed on Pelvis Through Meta-heuristically Optimised WFLC

Abstract

The estimation of lower trunk orientation and position during normal walking is relevant in clinical setting in order to improve the assessment of walking disorders. In this paper we introduce a new method for the estimation of the position of an Inertial Measurement Unit (IMU) placed on pelvis, during normal walking on a treadmill. The element of innovation is the use of a meta-heuristic optimisation process to estimate the optimal parameters of a Weighted Fourier Linear Combiner (WFLC) filter, which is designed to efficiently extract periodic/pseudo-periodic components of signals. The estimation of WFLC parameters was performed through an optimisation procedure based on the Artificial Bee Colony (ABC) algorithm, minimising the difference between the WFLC reconstructed position and the data coming from a sterophotogrammetry (SP) system. The WFLC weights obtained from the first set of data (training set), with different walking speeds, were then used to improve the estimation of multiple walking trials with the same measurement setup (test set). This approach allows to obtain useful clinical information using wearable, lightweight and low power consuming devices such as IMUs. This method has been validated through SP data, evaluating the Root Mean Square Error (RMSE) between the two system’s position estimations. The results show a global improvement of the position estimation over the three axes both during the training phase and the test phase. A low SD among the RMSEs in the test set, after the filter application, shows a good repeatability of the method over different trials at the same speed.