Classification of mental tasks using de-noised EEG signals

Abstract

The wsvelet based de-noising can bc eiiiployed ivi[li ihc combii.~ation of different kind of threshold parameters. tlwesliold operators. mother wavelets and timsliold rescaling methods. The central issue i.n wavelet bassd de-noising method is the selection of an appropriate ilircshold paraiiwters. If the tlmsliold is too small. the signal is still noisy but if it is too large. iiiiportant signal Cc;itiires might lost. This study will inyestigate the cfl'ectivcness of Tour Qpes of threshold parameters i.e. t lircshold selections based on Stein's Unbiased Risk Esliriiate (SURE,). Universal. Heuristic and Minimax. Auioregressiite Burg model with order sis is employed to cstmct rcle\*ani features froin the clean signals. These Icmrcs are classificd into five classes of mental tasks via :in iinificial neural nctwork. Tlie results show that the rate of correct classification varies with different tlucsliolds. From this study. it shows that the de-noised EEG signal tvitli heuristic threshold selection outperform lhe others. Soft tl~esholding procedure and sym8 as the mother U aixlcl arc adoptcd in this study.