Biomechanical Ambulatory Assessment of 3D Knee Angle Using Novel Inertial Sensor-Based Technique

Three-dimensional (3D) knee angle measurement is one of the key measures in human gait analysis. Inertial sensor capable measuring joint motion under unconstrained conditions a practical tool for clinical evaluation and rehabilitation. An inertial unit (IMU) consisting accelerometer gyroscope allows orientation 3D with an additional (i.e., magnetometer). However, ferromagnetic interference negatively affects performance magnetometer thus reduces accuracy. In this study, technique based on nonlinear autoregressive neural network exogenous inputs (NARX) presented to measure segmental during without use magnetometer. With IMUs attached thigh shank, angles long-distance treadmill walking were computed validated against optical analysis system as gold standard. Pseudo-integrator (PI) was also compared reference benchmarking. The learning capability NARX further assessed comparison complementary filter (CF) system. proposed model shown outperform PI biases between -3.5° -0.2°, root mean square errors 4.5° 2.5°. Results demonstrated providing accurate estimates while avoiding encountered systems incorporating magnetometer, suggesting that feasible computing long-term ambulatory measurements body segment angles.