Application of Wiener Deconvolution Model in P300 Spelling Paradigm

Spelling paradigm first introduced by Farwell and Donchin, is one of the brain computer interface (BCI) applications that enables paralyzed people to communicate with their environment. In such a problem, user needs to focus on the characters which are randomly flashed row or column-wise on the computer screen in a small period of time. The accuracy in spelling words is the main problem in this scheme and the duration of the correct prediction is quite important. The purpose of this work is twofold: to analyze a user specific response to a spelling paradigm system considering the optimal frequency bands for P300 detection, and secondly to investigate the classification performance for the perception of row and columnwise flashings in the spelling system. The preprocessing is performed with Wiener deconvolution model (WDM) and optimal filters for user specific system is constructed. The proposed algorithm is applied to dataset IIb of BCI competition 2003 and the words for training and testing sets are predicted with 100% accuracy after first 4 trials, as compared to other winning algorithms (100% accuracy in 5 repetitions) of the competition. Furthermore, our classification results show that perception to row and column flashings may differ considerably.