Evaluation of Visual Selective Attention by Event Related Potential Analysis in Brain Activity

This research proposes on the evaluation of different brain regions in visual selective attentional states by event-related potential (ERP) signals. Selective attention refers to the limitations of brain processing capacity in multiple simultaneous stimuli. Extracted epochs that are time-locked to stimuli onsets in each group are averaged for calculate the ERPs. In this paper, feature extraction by wavelet coefficients and time-shape features, feature selection by p-value and scattering criteria and classification by support vector machine (SVM) with Gaussian and polynomial kernels. Here, 5-fold cross validation use for result assessment. The results show that the best separation interval between responses is 100 to 400 msec by stepwise analysis method. The P3b peak on target stimuli is clearly observed in comparison with non-target stimuli. The target and non-target categories are separate with the average accuracy of 86.7% by scattering criteria and Gaussian SVM. The temporal and parietal regions of the brain have better performance in comparison with other regions. Additionally, there is no dominancy in the cerebral hemispheres. Therefore, this approach is a useful method in representation of brain activity in visual selective attentional state.