TK1115 : Using mexta-heuristic methods to improve the performance of emotion recognition in convolutional neural networks (CNN) baxsed on electroencephalogram signals (EEG)
Thesis > Central Library of Shahrood University > Electrical Engineering > MSc > 2026
Authors:
Abstarct: Abstract
Emotion recognition baxsed on electroencephalography (EEG) signals reflecting affective states is an important topic in computational neuroscience and artificial intelligence. Although Convolutional Neural Networks (CNNs) have a high capability for extracting complex features, they face challenges in architecture design and hyperparameter tuning, including their reliance on trial-and-error approaches, which lead to high computational cost and human bias.
In this study, a two-dimensional Convolutional Neural Network was designed for emotion recognition using the SEED dataset, and its hyperparameters were optimized using two mextaheuristic algorithms: the Grey Wolf Optimizer (GWO) and the Whale Optimization Algorithm (WOA). A cost function baxsed on validation accuracy and validation error was formulated within a five-fold cross-validation frxamework. Ten key hyperparameters, including the learning rate, activation functions, number of neurons, weight regularization, and batch size, were optimized. For each mextaheuristic optimizer, a population size of 15 and 10 iterations were considered.
The results showed that the proposed method improved the emotion recognition accuracy from 83% to 86%, while simultaneously reducing the number of trainable parameters, operational complexity, storage memory requirements, and runtime memory requirements. These findings demonstrate the effectiveness of the mextaheuristic approach in optimizing Convolutional Neural Networks and its applicability to other deep learning problems.
Keywords:
#_EEG signals #emotion recognition #brain signal processing #mextaheuristic algorithms #convolutional neural networks Keeping place: Central Library of Shahrood University
Visitor:
Visitor: