Multichannel surface EMG based estimation of bilateral hand kinematics during movements at multiple degrees of freedom

The paper proposes a method to estimate wrist kinematics from surface EMG signals for proportional and simultaneous control of multiple degrees of freedom (DOFs). The approach is based on the concurrent detection of surface EMG signals from forearm muscles and hand kinematics of both limbs during mirrored bilateral movements in free space which involve the simultaneous activation of wrist flexion/extension, radial/ulnar deviation and forearm pronation/supination. The estimation was based on one multilayer perceptron (MLP) neural network for each DOF. The three MLPs were trained to estimate angular displacements corresponding to the three DOFs. The average coefficient of determination between the true and the predicted angular displacement was 82.7 ± 2.9% (80.9 ± 3.4%) for flexion/extension, 75.0 ± 3.8% (72.6 ± 9.4%) for radial/ulnar deviation, 76.6 ± 11.8% (75.1 ± 11.7%) for pronation/supination for the ipsi-lateral (contra-lateral) hand. The scheme represents a step forward towards the simultaneous control of DOFs and thus a more natural prosthetic control.