Designing a spatial filter to improve SNR in two-class discrimination problems for BCI applications

The accuracy in classifying electroencephalographic (EEG) data in brain-computer interfaces (BCI) depends on the number of measuring channels, the amount of data used to train the classifier, and the signal-to-noise ratio (SNR). Of all those factors, the SNR is the hardest to adjust in real-life applications. For this reason, a spatial filter based on a linear minimum mean squared error (LMMSE) beamformer is proposed to increase the SNR of the EEG signals before they are passed to the classifier. For the special case of discriminating between two different neural events, the spatial filter is designed through the best discriminating hyperplane obtained from a Fisher´s discriminant analysis of the training data. A series of simulations show that the proposed spatial filter is effective in improving the mean performance of a given classifier under low SNR conditions, while optimal performance is achieved for a sufficiently large set of training data.