Categorizing visual objects; using ERP components

Paying attention to different pictures is related to complex information processing in the brain. Categorizing visual objects using the electroencephalogram (EEG) signal of subject along with paying attention to pictures, is properly possible. The aim of this paper is to analyze the mental signal in order to show the differences in cognitive patterns during paying attention to sets of different pictures. For this purpose, EEG signals which were recorded from 45 people were used. Brain signals are recorded over the on head using 8 active electrodes and based on standard 10-20. After the pre-processing, ERP signals were extracted into two classes according to attention to the human face and fruit images. Firstly, 4 types of features has been extracted from N170, P200, N200 and P300 components: (1) time features, (2) non-linear features, (3) statistical features and (4) frequency features. Then dimension of Properties were reduced by using different algorithms. New and innovative work in this paper is using various algorithms for reducing feature dimension such as t-test, t-SNE and kernel t-SNE and comparing their results with each other. Classification of 2 classes were done in order to recognize the differences using SVM and KNN classifiers. Secondly we reexamined this process by using combined features from multiple ERP components and obtained best result in this condition by t-SNE and SVM classifier with 85.5% accuracy.