Optimizing channel selection using glow swarm optimization for brain computer interface

Abstract

Brain Computer Interfacing BCI allows non muscular communication between a human and a computer by detecting a users intentions via brain signals eg with an Electroencephalogram EEG and translating them into control commands This is particularly useful for people affected by diseases that lead to the loss of muscular control such as Amyotrophic Lateral Sclerosis ALS brainstem stroke multiple sclerosis and especially for people who suffer from locked in syndrome BCIs can not only be applied for communication but also for the control of external devices such as a wheelchair for rehabilitation and mental state monitoring In most BCI the identification of pattern is based on a classification algorithm ie an algorithm that automatically estimates the class of data represented by a feature vector of the EEG Because of growing interest for EEG based BCI many published results are related to investigation evaluation of classification algorithms The BCI research has captured the commercial interests and also the potential for the application in military services A method that is used for improving the performance of the EEG based BCI is by using suitable channels This is owing to the fact that if it is noisy or redundant the channels are removed and the complexity of computation is decreased Also the use of huge number of channels will not be practical as in need a long time for being set up As the appropriate channels can be different from one subject to another a method to identify subject specific optimal appropriate channels is very important in the BCI application performance In this work the patterns of common spatial used for extraction of feature and also the method of selection of electrode used for the choosing of features The naïve Bayes classifier along with the k Nearest Neighbor kNN classifier has been used for classification of features newline