Emotion Recognition from Electroencephalogram (EEG) Signals Using a Multiple Column Convolutional Neural Network Model

Emotions are vital in human cognition and are essential for human survival. Emotion is often associated with smart decisions, interpersonal behavior, and, to some extent, intellectual cognition. From the recent literature on emotion recognition, we understand that the researchers are showing interest in creating meaningful "emotional" associations between humans and machines; there is a demand for accurate and scalable systems to detect human emotional states, as emotion recognition is needed to understand the mental status of such persons who cannot communicate their emotions, such as disabled people, mentally challenged persons, etc. Therefore, EEG signals provide a non-invasive method to identify the emotions of these disabled humans. The research community has recently been very interested in employing electroencephalography (EEG) for emotion classification since end-users have wearable EEG systems that may offer a portable, cheap, and straightforward technique for identifying emotions. Deep learning models have recently been extensively used to extract characteristics and recognize emotions from EEG recordings. Apart from that, various papers were reviewed in this research. This paper presented a multiple-column CNN network with a leaky ReLU activation function on the EEG brain wave and DEAP datasets. Multiple scalp electrode locations are used to collect EEG signals, and each electrode offers spatially unique data. The network can detect spatial correlations and extract characteristics that depict the spatial distribution of brain activity by employing a multiple-column CNN, which simultaneously processes signals from many channels. This makes it possible for the model to recognize emotions using the spatial information in EEG data. The result analysis was evaluated on different CNN models, and it was observed that an accuracy of 98.10% was achieved on the EEG brainwave dataset and 81% on the DEAP dataset. © 2024, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.