Comparison of Performance of Different Feature Extraction Methods in Detection of P300

The aim of this paper is to design a pattern recognition based system to detect the P300 component in the EEG trials. This system has two main blocks, feature extraction and clas-sification. In the feature extraction block, in addition to morphological features, some new features including intelligent segmentation, common spatial pattern (CSP) and combined features (CSP + Segmentation) have also been used. Two criteria were used for the feature evaluation. Firstly, a t-test has been applied. Secondly, each of these four groups of features was evaluated by a Linear Discriminant Analysis (LDA) classifier. Afterwards, the best set of features was selected by using Stepwise Linear Discriminant Analysis (SWLDA). In the classification phase, the LDA was used as a linear classifier. The algorithm described here was tested with dataset II from the BCI competition 2005. In this research, the best result for the P300 detection was 97.4%. This result has proven to be more accurate than the results of previous works carried out in this filed.