An Evolutionary Convolutional Neural Network Architecture for Recognizing Emotions from EEG Signals

Diagnosis and treatment of various emotions are crucial in modern life to prevent chronic emotional states and irreversible damage. By integrating the Internet of Things (IoT) and automated learning strategies within residential settings, it is now feasible to design intelligent environments capable of detecting and recognizing emotions induced by stress. To address the challenges of emotion analysis, this chapter presents a hybrid model that combines evolutionary convolutional neural network-based learning (evCNN) with emotion recognition. The primary objective of our research is to deploy hybrid learning techniques to diagnose different emotions within the context of the healthcare structure and the Internet of Medical Things (IoMT). In this study, we introduce the optimized gray wolf algorithm (opGWO) as an innovative method for determining the optimal parameters within the deep CNN architecture. The proposed architecture was evaluated for its ability to generalize emotion recognition using previously unseen EEG signals. Additionally, we reduced computational complexity by using a limited number of EEG channels to evaluate emotions. Our approach demonstrated a precision of over 90% in classifying emotions across two datasets.