Neural drive estimation using the hypothesis of muscle synergies and the state-constrained Kalman filter

We explore the hypothesis of muscle synergies to estimate the neural drive (movement intent) for upper extremity myoelectric prosthesis using the surface myoelectric signals. Commonly employed pattern classification systems have certain limitations, like inherent discrete nature, finite movement classes and limited degrees-of-freedom. We propose a novel framework based on the state space modeling and the hypothesis of muscle synergies. The problem is formulated in the state space framework in a novel way, where the movement intent is modeled as the hidden state of the system. A continuous stream of the movement intent (the hidden state) is estimated using the state-constrained Kalman filter. Preliminary experimental results also confirm the applicability of the proposed framework for estimation of movement intent.