FIGURE 3.

EEG signal processing. After performing an independent component analysis (ICA), ERSPs of each independent component (IC) were calculated for each trial using EEGLAB. Then, 1,500 ERSP values were obtained at each frequency at each time point in each IC (since there were 1,500 trials). We performed classification analyses using different class labeling depending on movements (i.e., 30 movements) and parameters (i.e., 24 directions, 7 distances, and 5 horizontal and vertical positions). Then, we performed an ANOVA using the 1,500 ERSP values at each frequency at each time point in each IC. If p > 0.05, ERSP values at that time point at that frequency in that IC were excluded from further analyses. If significant, post hoc analyses were performed to find significant pairs. An ERSPs at a particular time and frequency in an IC was selected as a feature for subsequent binary classifications (for significant pairs). After this procedure was completed with respect to all ICs, frequencies, and times, binary classifications using significant ERSPs were performed. ANOVA, analysis of variance; Freq, frequency; ERSP, event-related spectral perturbations.