Affective Computing on Machine Learning-Based Emotion Recognition Using a Self-Made EEG Device

Area of Science:

• Neuroscience
• Computer Science
• Affective Computing

Background:

• Emotion recognition is crucial for human-computer interaction.
• Electroencephalography (EEG) offers a direct measure of brain activity.
• Wearable EEG devices enable real-time affective state monitoring.

Purpose of the Study:

• To develop and evaluate a machine learning-based affective computing method for emotion recognition.
• To utilize a custom-designed wearable EEG device and wireless protocol.
• To assess the performance of different classifiers and feature extraction techniques.

Main Methods:

• Collected EEG signals from eight subjects using an eight-electrode placement (frontal and temporal lobes).
• Extracted features using six entropy measures.
• Classified emotions using Support Vector Machine (SVM), Multi-layer Perceptron (MLP), and 1D Convolutional Neural Network (1D-CNN).
• Evaluated both subject-dependent and subject-independent classification strategies.

Main Results:

• The highest average accuracies were 85.81% (subject-dependent) and 78.52% (subject-independent).
• Optimal performance was achieved using sample entropy and a 1D-CNN classifier.
• The T8 electrode position (right temporal lobe) was identified as the most critical for emotion classification.

Conclusions:

• The developed EEG-based affective computing method is feasible and efficient for emotion recognition.
• Wearable EEG systems show promise for real-world emotion detection applications.
• Specific EEG channels and entropy measures can significantly improve classification accuracy.