A Novel Approach for Upper Limbs Joint Angle Measurement Using Wearable IMU Sensors

This paper introduces a novel method for measuring upper limb joint angles using wearable IMU sensors and a Kalman filter algorithm. The proposed method estimates 3D joint orientations without relying on magnetometer data, addressing limitations associated with traditional motion capture systems. To assess the accuracy and reliability of IMU-based measurements, a comprehensive evaluation against the gold standard Vicon system is conducted. The agreement between IMU and Vicon measurements is assessed using various error metrics, including RMSE, MAE, NMBE, and R-2, followed by Mann-Whitney U tests with Bonferroni correction for statistical significance. Results indicate strong correlations between IMU and Vicon data (R-2 values: 0.890 to 0.974, R-2 > 0.7). RMSE (5.35 degrees to 12.07 degrees) and MAE (3.69 degrees to 7.01 degrees) values represent a small percentage of the range of motion for each joint. NMBE values are mostly close to zero, indicating minimal bias in IMU measurements. Notably, wrist flexion-extension exhibits a relatively higher NMBE (-0.839), suggesting a potential underestimation of angles in this plane of motion. Statistically significant differences (p < 0.05 with Bonferroni correction) are found for all degrees of freedom, with discrepancies ranging from 1.7% to 21% of the total dataset. Smallest differences are observed for wrist flexion-extension and elbow pronation-supination, while the largest occur in elbow flexion-extension and shoulder adduction-abduction. These findings suggest the proposed algorithm's reliability, as evidenced by strong correlation with the Vicon system and low error rates and biases for most degrees of freedom. However, larger population studies are warranted for further validation and refinement.