An Inertial-Based Upper-Limb Motion Assessment Model: Performance Validation Across Various Motion Tasks
Lin Meng, Mengjuan Chen, Baihan Li, Feng He, Rui Xu, Dong Ming
Abstract
Inertial measurement unit (IMU) has been widely applied in the development of upper-limb motion assessment model due to its portability, low cost, and theoretically unlimited workspace. However, most inertial-based models do not meet the accuracy requirement of clinical use. The errors mainly arise from the sensor measurement and “sensor-to-segment” misalignment. This article presented a novel inertial-based upper-limb motion assessment model with two layers, which enabled simultaneous estimation of the joint angles and positions. The sensor-to-segment dynamic calibration method was employed in the construction of the upper-limb kinematic model. The joint angles were calculated based on orientation transforms between adjacent body segments, while the joint positions were estimated using a forward kinematic chain. The model performance was evaluated with the terms of the Pearson correlation coefficient (PCC), offset, and root-mean-square error (RMSE) by comparing with results derived from the Vicon reference system. Effects of motion type and motion plane on estimation accuracy were further investigated with the use of a two-way analysis of variance (ANOVA). The results showed that the estimated joint angles and positions had a good agreement with the reference, and the RMSEs below 5.79° and 3.45 cm were, respectively, achieved. Both the motion type and motion plane had significant effects on the model performance, suggesting that the motion task effect should be considered in applications of inertial-based motion analysis methods. The proposed inertial model has great potential in upper-limb movement evaluation for patients.