Wavelet Filtering of the P300 Component in Event-Related Potentials Seyedehmina Ayoubian Markazi , Student Member, IEEE, S. Qazi , Student Member, IEEE, Lampros. S. Stergioulas , Member, IEEE , , Anusha Ramchurn , & David Bunce School of Information Systems, Computing and Mathematics, Brunel University, UK Centre for Cognition and Neuroimaging, Brunel University, UK Department of Psychology, Goldsmiths College, University of London, UK ,* a a c a b a b c Abstract— This paper presents an application of wavelet filtering to single-trial P300 component analysis. The objective of this study is to introduce a new method for analyzing the P300 component, when performing a given cognitive task, in this case, a two-choice reaction time task. The discrete wavelet transform with Daubechies wavelet is employed to detect the presence of P300 in individual trials. Wavelet filtering is applied to remove noise and unwanted frequency components from discrete wavelet transform (DWT) coefficients based on prior knowledge of event-related potentials (ERPs). The filtering mask is computed from the grand-average of wavelet coefficients over all participants. With this filtering, the P300 component is accurately localized in both time and scale. The findings suggest the procedure to have considerable potential for the analysis of time-series data in the behavioral neurosciences. Keywords—Age, Electroencephalography, EEG, ERP, P300, Wavelet transform, Wavelet filtering. I. INTRODUCTION Electroencephalography is the neurophysiologic measurement of the electrical activity of the cerebral cortex of the brain. The recorded brain activity is known as an electroencephalogram (EEG) or brain dynamics. EEG activity occurs continuously in both humans and animals; however, if EEG activity is recorded in relation to a specific stimulus, it is then referred to as an evoked related potential (ERP). Recently there has been a growing interest in brain dynamics, which provides valuable insights into a wide variety of neurological activities and disorders. One of the ERP components that is commonly investigated in behavioural neuroscience research is the P300. In cognition terms, P300 is considered to represent stimulus evaluation time (latency) and attention engagement (amplitude). It’s worth noting that as the P300 is a particularly large component, it lends itself to single–trial analysis. Since the discovery of P300 in 1965, a large body of research has been carried out to understand the related cognitive mechanisms and their underlying activities of the brain (Bashore & van der Molen, 1991). A whole range of techniques is available for the study of brain responses, which have found important users in several research fields such as evolutionary developments of the brain, aging, pathology and pharmacology. One of the applied research domains is the analysis of P300 ERP in the aging process (Basar, 2004). __________________ * Corresponding author.Tel: +44 (0) 1895265968 e-mail: seyedehmina.ayoubian@brunel.ac.uk EEG and ERP data are good examples of non-stationary signals, with varying frequency content. The time evolution of the amplitudes in single-trial ERPs does not allow the accurate retrieval of the frequency information of the signal. Spectral analysis can offer a more informative way to analyse ERPs. The Fourier transform, which is the most common spectral analysis tool and is almost universally used for stationary signal analysis, fails to provide any information about the time domain. Since the frequency content of ERPs is time dependent, time-frequency analysis is best suited for analysis of this type of signals. The Short-Time-Fourier Transform (STFT) maps a signal into a two-dimensional function of time and frequency. The STFT represents a compromise between the time and frequency content of a signal. However, the information it provides in either domain is limited, and this limitation is determined by the size of the window function which is zero-valued outside of some chosen interval. In this paper, the Wavelet transform (WT), a well-known time-scale analysis method, is employed. The major advantage of the WT is the use of variably-sized regions for the windowing operation. Wavelet analysis uses both long time windows enabling more precise estimation of low- frequency information, and shorter time windows for high- frequency information. As it provides time, scale (frequency) and amplitude information, with an all-round satisfactory resolution, it enables efficient multiresolution analysis (MRA). Thus, the variable time resolution of WT matches the structure of single-trial ERP signals and provides an efficient analysis of the non-stationary nature of transient signals such as ERPs. Proceedings of the 28th IEEE EMBS Annual International Conference New York City, USA, Aug 30-Sept 3, 2006 ThEP3.13 1-4244-0033-3/06/$20.00 ©2006 IEEE. 1719