Article Preview
TopDifferent methods of signal classification have been tested with the P300-BCI Speller including Stepwise Linear Discriminant Analysis (SWLDA) (Krusienski, Sellers, McFarland, Vaughan, & Wolpaw, 2008), Support Vector Machine (SVM) (Kaper, Meinicke, Grossekathoefer, Lingner, & Ritter, 2004), Independent Component Analysis (ICA) (Li, Sankar, Arbel, & Donchin, 2009), Matched Filter (Serby, Yom-Tov, & Inbar, (2005), and Wavelet Analysis (Bostanov, 2004). Krusienski, Sellers, Cabestaing, Bayoudh, McFarland, Vaughan, and Wolpaw (2006) investigated the techniques mentioned above and concluded that SWLDA was the most accurate and practical processing method for P300 speller. However, SWLDA and other techniques share the same drawback. They need to average several trials to remove the background noises and enhance the magnitude of P300 response before applying the P300 classifier on EEG signal. This need for averaging slows the communication process of the P300 BCI. It would clearly be advantageous to employ a method that would allow reliable detection of P300 in a single trial. Our goal in the present study is to develop an algorithm for the single trial detection of P300 response.