Neuropsychology Identification of Emotion and Attention Using Multiple Hybrid 2D Convolutional Neural Networks and Gated Recurrent Units

Neuropsychology studies the relationship between the structure and function of the brain with human psychological behavior. One of its uses in neuromarketing is to evaluate the response of marketing products based on an analysis of brain activity through an Electroencephalogram (EEG). Neuropsychology captures EEG variables such as emotion, attention, and enjoyment. It is one or a combination. Using multi variables is an advantage in representing more comprehensive neuromarketing assessments. Each variable has its characteristics, and using multi variables separately needs multi networks or separate networks. This paper proposed multiple networks. Each variable EEG signal is filtered using Wavelet according to the frequency band that characterizes it. Then, EEG signals from multiple channels were extracted using two-dimensional Convolutional Neural Networks (2D-CNN) as spatial directions, then identified using Recurrent Neural Networks (RNN) as temporal. Gated Recurrent Unit (GRU) is one technique to arrange memory in RNN applications. We call the proposed method multiple hybrid 2D CNN-GRU. Experiments prove that the use of multiple networks outperformed single networks. In addition, it is seen that 2D-CNN performs better than 1D-CNN in spatial extraction. The choice of learning weight correction technique also affects performance, so the proposed method provided an accuracy of 91.7% and a Loss value of 0.2807.