Single-trial evoked potential estimation using wavelets

In this paper we present conventional and translation-invariant (TI) wavelet-based approaches for single-trial evoked potential estimation based on intracortical recordings. We demonstrate that the wavelet-based approaches outperform several existing methods including the Wiener filter, least mean square (LMS), and recursive least squares (RLS), and that the TI wavelet-based estimates have higher SNR and lower RMSE than the conventional wavelet-based estimates. We also show that multichannel averaging significantly improves the evoked potential estimation, especially for the wavelet-based approaches. The excellent performances of the wavelet-based approaches for extracting evoked potentials are demonstrated via examples using simulated and experimental data.