Wavelet-based Processing of EEG Data for Brain-Computer Interfaces

Brain-Computer Interfaces based on non-invasive electroencephalographic (EEG) signals were recently made practical through sophisticated algorithms and clever systems, in such a way that the dream of effortlessly translating volition into action is coming true, albeit in a limited way. However, a low signal-to-noise ratio and the presence of frequent artefacts, such as eye blinks, contaminate the recordings and make the recognition of the underlying mental processes difficult. In this study, a novel wavelet-based signal processing technique, Continuous Wavelet Regression, has been applied to refine EEG data in a well-known setting. The recordings of spontaneous (i.e., asynchronous) signals of subjects performing highly different cognitive tasks have been processed by our algorithm, and then analyzed and classified, obtaining very promising results as compared with those obtained by previous studies.