Machine learning techniques to classify emotions from electroencephalogram topographic maps: A systematic review.

Area of Science:

• Neuroscience and Artificial Intelligence
• Brain-Computer Interfaces

Background:

• Facial expressions are common for emotion recognition, but electrical brain signals offer higher data integrity.
• Electroencephalogram (EEG) devices capture neural activity non-invasively.
• EEG signals can be converted into graphical EEG topographic maps (ETMs).

Purpose of the Study:

• To identify and map machine learning techniques for emotion recognition from EEG topographic maps.
• To provide a state-of-the-art overview of current methods.
• To guide future research directions in affective computing.

Main Methods:

• Systematic literature review following PRISMA guidelines, with searches up to July 2025.
• Inclusion of 14 publications meeting specific criteria.
• Analysis of machine learning models, datasets, feature extraction, and EEG signal conversion methods.

Main Results:

• Identified machine learning techniques including Support Vector Machines (SVM), Convolutional Neural Networks (CNNs), Lightweight CNNs (LCNN), VGG-16, Bidirectional Long Short-Term Memory networks (Bi-LSTM), Residual Networks (ResNet), and Multilayer Perceptrons (MLP).
• Summarized correlations involving emotional datasets, feature extraction, and EEG to ETM conversion.
• Discussed classification accuracy across subject-dependent, subject-independent, transfer learning, and cross-subject approaches.

Conclusions:

• Machine learning, particularly deep learning models like CNNs and LSTMs, shows promise for emotion recognition using EEG topographic maps.
• Further research is needed to optimize subject-independent and transfer learning approaches for robust emotion recognition.
• EEG topographic maps provide a valuable representation for analyzing neural correlates of emotion.