Effect of walking variations on complementary filter based inertial data fusion for ankle angle measurement

Abstract

A key problem on the measurement of lower-limb joint angles using inertial sensors is drift resulted in error accumulation after time integration. Several types of methods have been proposed to eliminate the drift. Among these methods, complementary filter-based sensor fusion algorithms are widely used in real-time applications due to its efficiency. Results from existing studies have shown that the performance of methods is relevant to walking speed. However, factors of walking variation have not been explored. This study first systematically investigated the walking variation factors and their effects on the accuracy of a proposed sensor fusion method during treadmill walking. Ten able-bodied participants participated in the experiment and walked on a treadmill with three different speeds (0.5, 1.0 and 1.5 m/s). A 12 camera Vicon motion capture system was used as the reference. The accuracy of the proposed method was evaluated in terms of the root-mean-square errors (RMSE), offsets and Pearson's correlation coefficients (PCC) in phases of a normalised gait cycle. A general linear model of analysis of variance (ANOVA) was used to analyze the factors including treadmill speed and gait phases. Results showed both factors had a significant influence on the RMSE, and only the treadmill speed had a significant influence on the offset. It provides an insight to improve the complementary filter-based method in future work.