AbstractMany of the brain-computer interface (BCI) systems depend on the user’s voluntary eye movements. However, voluntary eye movement is impaired in people with some neurological disorders. Since their auditory system is intact, auditory paradigms are getting more patronage from researchers. However, lack of appropriate signal-to-noise ratio in auditory BCI necessitates using long signal processing windows to achieve acceptable classification accuracy at the expense of losing information transfer rate. Because users eagerly listen to their interesting stimuli, the corresponding classification accuracy can be enhanced without lengthening of the signal processing windows. In this study, six sinusoidal amplitude-modulated auditory stimuli with multiple message frequency coding have been proposed to evaluate two hypotheses: 1) these novel stimuli provide high classification accuracies (greater than 70%), 2) the novel rhythmic stimuli set reduces the subjects’ fatigue compared to its simple counterpart. We recorded EEG from nineteen normal subjects (twelve female). Five-fold cross-validated naïve Bayes classifier classified EEG signals with respect to power spectral density at message frequencies, Pearson’s correlation coefficient between the responses and stimuli envelopes, canonical correlation coefficient between the responses and stimuli envelopes. Our results show that each stimuli set elicited highly discriminative responses according to all the features. Moreover, compared to the simple stimuli set, listening to the rhythmic stimuli set caused significantly lower subjects’ fatigue. Thus, it is worthwhile to test these novel stimuli in a BCI experiment to enhance the number of commands and reduce the subjects’ fatigue.Significance StatementAuditory BCI users eagerly listen to the stimuli they are interested in. Thus, response classification accuracy may be enhanced without the need for trial lengthening. Since humans enjoy listening to rhythmic sounds, this study was carried out for introducing novel rhythmic sinusoidal amplitude-modulated auditory stimuli with multiple message frequency coding. Our results show that each stimuli set evoked reliably discriminative responses according to all the features, and rhythmic stimuli set caused significantly lower fatigue in subjects. Thus, it is worthwhile to test these novel stimuli in a BCI study to increase the number of commands (by NN permutations of just N message frequencies) and reduce the subjects’ fatigue.
Neurofeedback training of control network improves SSVEP-based BCI performance in children
