Multi-Class Brain Computer Interface Using Steady-State Somatosensory Evoked Potentials

Steady-state somatosensory evoked potential (SSSEP) is response of electroencephalogram to mechanical or electrical vibratory stimulation to human body such as a finger and sole, which is synchronized with the stimulus frequency in the range of 15-35 Hz. Brain computer interface based on SSSEP (SSSEP-BCI) has been proposed by exploiting the modulation of SSSEP associated with selective attention on one of two tactile stimuli. However, the number of options in the most of conventional SSSEP-BCIs has been limited to two for expressing user’s intentions. This study thus aimed at improving the performance of SSSEP-BCI by increasing the number of intentional options from two to four through the addition of new mental task “Shiritori” to the conventional selective attention task. The highest overall discrimination accuracy reached 78.3 % and 56.3 % for the proposed three-class and four-class SSSEP-BCIs, respectively. The highest discrimination accuracies for the proposed BCIs were obtained when the amplitude of alpha waves in the Shiritori task was combined with SSSEP as features for discrimination using the support vector machine, suggesting that the validity of the mental task for SSSEP-BCI. These results demonstrate the feasibility of a multi-class SSSEP-BCI which has more than three classes as intentional options.