Figure 6

Across-subject prediction of EMG activity. Example segment (~100 s) of predicted EMG signals (red traces) made by the MLP30 artificial neural network for anterior deltoid (AntDelt), brachioradialis (BrachRad), and extensor carpi radialis (ECR). The network was trained using loaded movement data from one subject and was then used to predict muscle activity patterns in a different test subject. Top trace indicates the grip force and black traces in lower three panels indicate the actual (recorded) EMG signals in the test subject. The coefficient of determination (R²) between predicted and actual EMG is shown for each muscle.