sfs algorithm

The present study aimed at scheming a novel method of detecting target and non-target signals through selection of appropriate and timely chronic intervals of VEP signal leading to increasing the accuracy of data classification and decreasing the number of features. The suggested method was employed on the P300-Speller databases of the BCI2005 competitions and the data recorded by Hoffman et al. using effective and specified channels and SWLDA classifier.The methods available for determining the P300 signals are within a specified range of about 1 second after each stimulation. To this end, we first outlined the time range of the various components of visual Evoked potential including N20, P50, N100, N170, P300, N400 based on the results obtained from the physiologically-based articles. Then, the time intervals were scored by F-Score and the percentages of correct classifications. The most important and effective components of the VEP were selected by SFS Algorithm  using the SWLDA classifier and the functions of the optimal combinations were compared with the total length of the signal utilizing two other classifiers namely Bayesand K_NN in order to confirm the functionality of the method. The findings, based on the results obtained from ten subjects, indicated that the most important components for detecting target and non-target signals include P300, N100, and N400 respectively. The method suggested here proved to improve the accuracy of output detection by 3.95%.