Surface EMG can collect physiological signals of muscle activity to infer the muscle movement or even the limb kinematics. It is very useful for developing more advanced wearable exoskeletons or rehabilitation robotics systems. This can be applied to be used for the upper limb amputation patients. Current research already demonstrates its effectiveness in limb motion tracking using various deep learning models, and also demonstrates its effectiveness for the real-time limb motion tracking system interacting with the robotics arm. However, the effectiveness of using sEMG for more generalized subjects or motions is less validated and developed.
In the EEG motion prediction, the continuous motion estimation often adopts Kalman filtering for more generalized subjects and motions. Inspired by the Kalman-based motion estimation strategy, this master's thesis aims to develop a forearm motion tracking system using the deep-learning and Kalman filtering methods to achieve real-time, generalized subjects, and generalized motion tracking.