EEG signal processing methods for BCI applications

Brain-computer interface (BCI) is a communication system that translates brain activity into commands for a computer or other digital device. The majority of BCI systems work by reading and interpreting cortically-evoked electro-potentials (“brain waves”) via an electroencephalogram (EEG) data. The EEG data is inherently complex. The signals are non-linear, non-stationary and therefore difficult to analyze. After acquisition, pre-processing, feature extraction and dimensionality reduction is performed, after witch machine learning algorithms can be applied to classify the signals into classes, where each class corresponds to a specific intention of the user. BCI systems require correct classification of signals interpreted from the brain for useful operation. This paper reviews our proposed methods for EEG signal processing and classification, which include Wave Atom transform, use of nonlinear operators, class-adaptive denoising using Shrinkage Functions and real time training of Voted Perceptron artificial neural networks.