Maximizing information transfer rates in an SSVEP-based BCI using individualized Bayesian probability measures

Successful brain-computer interfaces (BCIs) swiftly and accurately communicate the user´s intention to a computer. Typically, information transfer rate (ITR) is used to measure the performance of a BCI. We propose a multi-step process to speed up detection and classification of the user´s intent and maximize ITR. Users randomly looked at 4 frequency options on the interface in two sessions, one without and one with performance feedback. Analysis was performed off-line. A ratio of the canonical correlation analysis (CCA) coefficients was used to construct a Bayesian probability model and a thresholding method for the ratio of the posterior probability of the target frequency over maximal posterior probability of non-target frequencies was used as classification criteria. Moreover, the probability thresholds were optimized for each frequency, subject to maximizing the ITR. We achieved a maximum ITR of 39.82 bit/min. Although the performance feedback did not improve the overall ITR, it did improve the accuracy measure. Possible applications in the medical industry are discussed.