3D upper limbs tracking through inertial sensors: calibration, methodology, and validation

In home-based motor rehabilitation, bio-feedback and remote monitoring are essential to maintain patient motivation and ensure appropriate monitoring of training effects. Wearable sensors can be exploited to provide this feedback both to the patient and to the therapist. This paper presents an algorithm for the real-time extraction of shoulder and elbow joint angles from three inertial measurement units (IMUs). The five angles extracted are defined according to the International Society of Biomechanics (ISB) convention. The algorithm includes a two-step calibration procedure (N-pose and T-pose), which allows placing the IMUs on the limbs freely, without the need to use the magnetometer, and the real-time computation of the joint angles. The algorithm was validated on the arm of the TIAGo robot by comparing the joint angles measured by the robot encoders against the joint angles derived from the IMU sensors. Results show that the presented method is accurate (the maximum RMSE is 4.96◦ for the pronation-supination angle) and robust to different IMUs positioning.