Objective. The current study evaluates an auditory brain-computer interface (BCI) protocol that allows users to mentally choose among 6 options. Methods. The protocol is based on an oddball P300 paradigm. To reduce mental workload, we introduce a change in the typical oddball paradigm task: Instead of passively counting the presented target auditory stimuli, we ask participants to simply mentally repeat them when they occur. In the study, ten healthy participants carried out two calibration sessions respectively with traditional mental count and with the proposed mental repetition and then three free item selection sessions using mental repetition. A comparison has been conducted between off-line count and mental repetition classification accuracies achieved by each participant during the calibration sessions. The mental workload difference between the count and repeat calibration sessions of each participant was evaluated by computing alpha (at Po8) and theta (at Fz) spectral power density (SPD) curves. Results. Nine out of ten participants showed a higher mental workload with the traditional mental count. The repeat activity was preferred by 8 out of 10 participants. The comparison conducted between off-line count and mental repetition classification accuracies shows a slightly worse average behavior for the repeat protocol in the first 8 averaged trials. With the proposed protocol we got an average on-line item-selection information transfer rate (ITR) over 2 trials of 2.35 bits min-1, and an average on-line accuracy over 10 trials of 81.7%. Conclusions. Although off-line classification based on mental count data got slightly better results, the proposed auditory BCI protocol with mental repetition achieved an on-line performance similar to the traditional counting oddball paradigm task but with a lower mental workload. Significance. The results obtained with healthy subjects suggest that the proposed protocol can be a simpler alternative to the mental count, with comparable performance and lower mental load.
