Wavelet Packet Sub-band Based Classification of Alcoholic and Controlled State EEG Signals

摘要

Electro-encephalogram (EEG) is one of the most practiced signals in brain computer interface systems. Several distinct EEG patterns have been analyzed in identifying physiological and psychological states. Work presented here focuses on classification of EEG patterns for alcoholic and controlled states. Third level sub-band energy features are generated for either classes using multi-resolution wavelet packet transformation. A well-known support vector classifier is employed to segregate these features in two well defined classes. Experimental results show significant improvement over wavelet tree feature extraction. Cross-validation tests confirm the greater classification accuracy for proposed technique.