The application of particle filters in single trial event-related potential estimation

Mohseni, Hamid Reza, Nazarpour, Kianoush ORCID: https://orcid.org/0000-0003-4217-0254, Wilding, Edward Lewis ORCID: https://orcid.org/0000-0002-9495-1418 and Sanei, Saeid 2009. The application of particle filters in single trial event-related potential estimation. Physiological Measurement 30 (10) , pp. 1101-1116. 10.1088/0967-3334/30/10/010

Full text not available from this repository.

Abstract

In this paper, an approach for the estimation of single trial event-related potentials (ST-ERPs) using particle filters (PFs) is presented. The method is based on recursive Bayesian mean square estimation of ERP wavelet coefficients using their previous estimates as prior information. To enable a performance evaluation of the approach in the Gaussian and non-Gaussian distributed noise conditions, we added Gaussian white noise (GWN) and real electroencephalogram (EEG) signals recorded during rest to the simulated ERPs. The results were compared to that of the Kalman filtering (KF) approach demonstrating the robustness of the PF over the KF to the added GWN noise. The proposed method also outperforms the KF when the assumption about the Gaussianity of the noise is violated. We also applied this technique to real EEG potentials recorded in an odd-ball paradigm and investigated the correlation between the amplitude and the latency of the estimated ERP components. Unlike the KF method, for the PF there was a statistically significant negative correlation between amplitude and latency of the estimated ERPs, matching previous neurophysiological findings*.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Psychology Engineering Cardiff University Brain Research Imaging Centre (CUBRIC)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > T Technology (General)
Publisher:	Institute of Physics
ISSN:	0967-3334
Last Modified:	07 Jun 2023 06:24
URI:	https://orca.cardiff.ac.uk/id/eprint/31220

Citation Data

Cited 22 times in Scopus. View in Scopus. Powered By Scopus® Data

Actions (repository staff only)

Edit Item