EEG rhythm based emotion recognition using multivariate decomposition and ensemble machine learning classifier.

Area of Science:

• Cognitive Neuroscience
• Affective Computing
• Human-Computer Interaction (HCI)

Background:

• Electroencephalogram (EEG) signals show promise for recognizing human emotions.
• Effective computing aims to enhance human-computer interaction by enabling computers to understand emotions.
• Automated emotion recognition from multichannel EEG signals is a key research area.

Purpose of the Study:

• To propose a novel approach for automatic emotion classification using multichannel EEG signals.
• To investigate the efficacy of EEG multichannel rhythmic features combined with ensemble machine learning (EML) classifiers.
• To evaluate the proposed method using leave-one-subject-out cross-validation (LOSOCV).

Main Methods:

• Multivariate fast iterative filtering (MvFIF) was employed to analyze EEG rhythm sequences (delta, theta, alpha, beta, gamma).
• Extracted features included Hjorth parameters and entropy measures from multichannel EEG rhythms.
• Feature selection was performed using the minimum redundancy maximum relevance (mRMR) approach.
• Ensemble machine learning classifiers, including subspace K-nearest neighbor (SS KNN), were utilized.

Main Results:

• The proposed method achieved high classification accuracy, ranging from 93.5% to 99.8%, particularly with gamma rhythm multichannel features and EML-based SS KNN.
• Comparisons with Support Vector Machine (SVM) and Artificial Neural Network (ANN) demonstrated the effectiveness of EML.
• Analysis of multi-class emotions using an ensemble-based bagging tree on gamma rhythm showed promising results.

Conclusions:

• Multichannel rhythmic features derived from EEG data, especially gamma rhythms, offer a powerful approach for automated emotion recognition.
• Ensemble machine learning classifiers provide a robust framework for high-accuracy emotion classification from EEG.
• This research presents a novel solution for analyzing multichannel rhythm-specific features in EEG data for affective computing.