Feature extraction and classification of EEG signals for mapping motor area of the brain

This paper presents the study of open source electroencephalogram (EEG) data from 30 subjects performing actual motor tasks, for localizing brain motor areas responsible for the tasks. The extracted features from independent component analysis (ICA) of the EEG data are Gaussian weighted to obtain feature vectors. Two dimensional scalp maps are used for task based selection of features belonging to the primary and sensory motor regions of the brain. The final feature vectors thus obtained are given as input to two classifiers, viz. linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA). Classification using LDA gives localization accuracies of 68.42% for right fist movement, 67.16% for left fist movement and 84.40% for both feet movement respectively. The corresponding classification accuracies for QDA were 92.98% for right fist movement, 70.15% for left fist movement and 98.58% for both feet tasks respectively. The average accuracy for motor task classification is 73.33% for LDA and 87.24% for QDA.