Fig 1.
Subject and the position of the flickering circles on LED monitor.
Vertical and horizontal field of view were 23.48 and 39.42 degrees respectively. Diameter of the circles corresponded to 3.94 degrees. Circle centers were positioned at 6.92 degrees from the central fixation point.
Fig 2.
A: Offline, B: Online.
Fig 3.
Mean accuracies (%) (N = 24) vs. stimuli length–offline task.
*: p<0.05, **: p<0.01, ***: p<0.001 (FDR corrected). Comparisons (t-test) between the classifiers.
Table 1.
Two-way repeated measures ANOVA table for mean accuracies in the offline task depending on classifier and stimuli length (i.e. duration).
Fig 4.
Median power spectrum of all subjects and the power spectrum of a representative subject (Subject 18) depending on the frequency (F) of the circle with attentional focus.
Left: f = 15 Hz, Right: f = 12 Hz, Down: f = 8.57 Hz, Up: f = 5.45 Hz. flickering frequency (f) in the offline task. Power was averaged across trials. Each single trace represents one of the 60 EEG channels.
Fig 5.
Median power topography of all subjects (A) and the power topography of a representative subject (Subject 18) (B) in different frequency bands depending on the flicker frequency (F) of the circle with the attention focus in the offline task. Left: f = 15 Hz, Right: f = 12 Hz, Down: f = 8.57 Hz, Up: f = 5.45 Hz. Power was averaged across trials.
Fig 6.
Mean accuracies (%) under different perturbations (N = 24)–online tasks (3 s).
*: p<0.05, **: p<0.01, ***: p<0.001 (FDR corrected). Comparisons (t-test) between the perturbations.
Table 2.
Two-way repeated measures ANOVA table for mean accuracies in the online task depending on classifier and perturbations.
Fig 7.
Median power spectrum of all subjects and the power spectrum of a representative subject (Subject 18) depending on the online conditions: Control, Speaking, Thinking, and Listening.
Power was averaged across trials. The activity relates to all flickering frequencies. Each trace represents one of the 60 EEG channels.
Table 3.
One-way ANOVA table for power depending on perturbation.
Fig 8.
Median power topography of all subjects (A) and the power topography of a representative subject (Subject 18) in different frequency bands (B) depending on the perturbation condition in the online tasks. Power was averaged across trials and all target frequencies.
Fig 9.
Pearson correlation between relative high alpha (10–13 Hz) power and the Naïve Bayes classifier performance in the Listening condition across subjects.
A–Topography of the correlations. Black dots represent the channels that belong to the negative cluster (p = 0.02, 1000 permutations). B—An exemplary scatter plot for one of the channels (C4) from the significant cluster.