EEG-Based Emotion Intensity Recognition using Machine-Learning and CNN-Ensemble Models.

Area of Science:

• Neuroscience
• Computer Science
• Psychology

Background:

• Emotion intensity recognition is crucial for mental state understanding and human-computer interaction.
• Electroencephalogram (EEG)-based spectrogram analysis shows potential for emotion classification, but intensity recognition remains difficult.

Purpose of the Study:

• To propose and evaluate a methodology combining EEG feature extraction and machine learning for accurate emotion intensity recognition.

Main Methods:

• Collected EEG signals from 20 participants during a semi-controlled experiment.
• Extracted time-domain, frequency-domain, and spectrogram features from EEG signals.
• Applied machine learning classifiers including SVM, RF, XGBoost, LGBM, and hybrid CNN models, evaluated with 10-fold cross-validation.

Main Results:

• The CNN+SVM model achieved high performance with accuracy, precision, sensitivity, and specificity of 0.996, and a kappa coefficient of 0.994.
• The CNN+RF model, optimized for subject-independent prediction, yielded an accuracy of 0.649 and a kappa coefficient of 0.298.

Conclusions:

• The proposed methodology effectively classifies emotion intensity using EEG and machine learning.
• This framework has clinical relevance for objective emotion intensity recognition in mental health assessments, stress management, and affective computing.