Optimizing spatial filters for robust EEG single-trial analysis

The common spatial pattern (CSP) provides a clearer image of brain activity in single-trial analysis and improves signal-to-noise ratio in conducting multichannel electroencephalogram (EEG) recordings. CSP is a popular method in brain-computer interface (BCI) research. BCI systems promise developments in usability, information transfer and robustness for which modern machine learning and signal processing techniques have been instrumental. CSP produces a data-driven supervised decomposition of the signal parameterized by a certain matrix and is used to analyze multichannel data based on recordings from two conditions. CSP filters maximize the variance of the spatially filtered signal under one condition while minimizing it for another condition and can be applicable to band-pass filtered signals in order to yield an effective discrimination of mental stats that are characterized by ERD/ERS (event-related desynchronizationevent-related synchronization) effects. CSP can be used to extract general discriminative spatio-temporal structure for multivariate data streams beyond EEG.