ERP classification using empirical mode decomposition
Nitin Williams, Ian P. Daly, Slawomir J. Nasuto, Douglas Saddy, Kevin Warwick · CentAUR (University of Reading) · 2009
Empirical Mode Decomposition (EMD) is applied to the problem of ERP classification for BCI. Artificial ERPs are created using a Neural Mass model assuming the additive model of ERP generation at different SNRs. Features are extracted from an EMDbased reconstruction and a Support Vector machine is applied to classify the ERPs. The classification method produced classification accuracies significantly better than chance for SNRs between 0 dB and -20 dB. Further, the EMD-based method produced significantly better classification than the conventional averaging method for SNR of -5 dB, -10 dB and -15 dB.