Computational Intelligence for Pattern Recognition in EEG Signals

Electroencephalography (EEG) captures brain signals from Scalp. If analyzed and patterns are recognized properly this has a high potential application in medicine, psychology, rehabilitation, and many other areas. However, EEG signals are inherently noise-prone, and it is not possible for human to see patterns in raw signals most of the time. Application of appropriate computational intelligence is must to make sense of the raw EEG signals. Moreover, if the signals are collected by a consumer grade wireless EEG acquisition device, the amount of interference is ever more complex to avoid, and it becomes impossible to see any sorts of pattern without proper use of computational intelligence to discover patterns. The objective of EEG based Brain-Computer Interface (BCI) systems is to extract specific signature of the brain activity and to translate them into command signals to control external devices or understand human brains action mechanism to stimuli. A typical BCI system is comprised of a Signal Processing module which can be further broken down into four submodules namely, Pre-processing, Feature Extraction, Feature Selection and Classification. Computational intelligence is the key to identify and extract features also to classify or discover discriminating characteristics in signals. In this chapter we present an overview how computational intelligence is used to discover patterns in brain signals. From our research we conclude that, since EEG signals are the outcome of a highly complex non-linear and non-stationary stochastic biological process which contain a wide variety of noises both from internal and external sources; thus, the use of computational intelligence is required at every step of an EEG-based BCI system starting from removing noises (using advanced signal processing techniques such as SWTSD, ICA, EMD, other than traditional filtering by identifying/exploiting different artifact/noise characteristics/patterns) through feature extraction and selection (by using unsupervised learning like PCA, SVD, etc.) and finally to classification (either supervised learning based classifier like SVM, probabilistic classifier like NB or unsupervised learning based classifiers like neural networks namely RBF, MLP, DBN, k-NN, etc.). And the usage of appropriate computational intelligence significantly improves the end results.