Investigation of the most appropriate mother wavelet for characterizing imaginary EEG signals used in BCI systems

Feature extraction is a very challenging task, since choosing discriminative features directly affects the recognition rate of the brain computer interface (BCI) system. The objective of this paper is to investigate the effect of mother wavelets (MWs) on classification results. To this end, features were extracted from 3 different datasets using 12 MWs, and then the signals were classified using 3 classification algorithms, including k-nearest neighbor, support vector machine, and linear discriminant analysis. The experiments proved that Daubechies and Shannon were the most suitable wavelet families for extracting more discriminative features from imaginary EEG/ECoG signals.