Fig. 4

Representative results from one subject during offline experiment. Figure shows the individual finger force estimation when applying the LDA (a) and the CSP-PE (b) using a subset of 10 sEMG channels selected as a regular grid as input for the regression. The task for the subject was to activate the fingers individually and selectively, one at a time. The dashed gray line is the recorded finger force. The continuous red line is the estimated force for the finger that was supposed to be active in the task (active phase), while the continuous black line is the estimated force for the fingers that should have been relaxed (silent phase, false activation). The vertical lines delineate the active phases for different fingers. The quality of tracking is similar for the two methods, in this specific example, but the CSP-PE generated less false activations. For example, compare the false activation of the little finger in LDA vs. CSP-PE during the active phase of the index finger