GAN-ACNN: a design decision-making algorithm based on EEG signals from different brain regions

Abstract

Decision-making is an integral part of an individual's life. Due to the small amount of data in the art design decision dataset, we increase the amount of data using data augmentation. However, different brain regions have distinct effects on the accuracy of art design decisions, so we divide the brain into four areas and investigate the influence of each on the results. We collect electroencephalogram (EEG) data from 16 subjects, apply a band-pass filter to filter it, and then feed it into a generative adversarial network (GAN) for data augmentation. The augmented EEG data is input to a convolutional neural network with a self-attention mechanism (ACNN). The experimental results show that the degree of influence of different brain regions is: right frontal lobe > right parietal-temporal-occipital lobe > left frontal lobe > left parietal-temporal-occipital lobe. In view of this, we find the six optimal channels for art design decisions, and their prediction results are better than those of all channels. In addition, our GAN-ACNN model achieves an average accuracy of 93.51%, which is an effective method. Finally, we also classify the DEAP dataset to evaluate the robustness of the model.