Fig. 4: Decoding based on electromyographic (EMG) signals.

To measure the contribution of potential muscular activity to syllable decoding, we computed a classifier based solely on EMG data from the right zygomaticus major and the orbicularis oris, separately for offline and online sessions. a CV accuracies based on EMG (dotted lines) and EEG (solid line) data acquired during the offline (blue) and online (salmon) sessions. Error bars indicate the standard error of the mean. b Training slopes calculated by fitting a linear model across the 5 training days, separately for EEG (solid filling) and EMG (striped filling) and the two sessions (offline: blue, online: salmon). Boxes represent the interquartile range (IQR), with the horizontal line indicating the median, and whiskers extending to data points that are within 1.5× the IQR from the upper and lower quartile. Individual points represent data from a single participant (n = 15). c Correlation between the learning slope representing the behavioral improvement in BCI-control and the slope obtained considering the CV-accuracy for the EMG-online dataset (n = 15). d Histogram representing for each frequency, the average features’ weights obtained considering the EMG activity recorded during the offline session (the histogram is to be qualitatively compared with Fig. 2b). Significance is denoted with * for p < 0.05.