Human scalp EEG processing: Various soft computing approaches

Presently high density EEG systems are available at affordable cost, with which the data dimension has gone up considerably. For efficient computation of this high-dimensional data, various soft computing paradigms are receiving increasing attention. In this survey we have identified certain soft computing techniques (by soft computing techniques we mean computational techniques that take into account the inherent uncertainties in the data and/or in the computing model) for pattern recognition/data mining, such as, neural networks, fuzzy logic, evolutionary computation, statistical discrimination and Bayesian inference, which have turned out to be particularly useful in processing human scalp EEG. Wherever possible results of comparative studies among various techniques have been presented. Analyses of EEG for various feature extraction are exceedingly challenging pattern recognition tasks. This survey has shown that on an average the artificial neural networks and Bayesian approaches have emerged more successful in EEG analysis than the other soft computing paradigms. For readability the paper has been kept as little technical as possible. Large number of references have been listed to aid searching for the technical details.