An innovative EEG-based emotion recognition using a single channel-specific feature from the brain rhythm code...

Area of Science:

• Neuroscience
• Bioinformatics
• Machine Learning

Background:

• Emotion recognition is crucial for brain-computer interface (BCI) applications in intelligent healthcare.
• Current methods often require complex data processing and extensive hardware.

Purpose of the Study:

• To propose an innovative approach for electroencephalography (EEG)-based emotion recognition inspired by bioinformatics.
• To reduce the complexity and data requirements for practical BCI emotion recognition systems.

Main Methods:

• Utilized brain rhythm code features (δ, θ, α, β, γ) inspired by genetic code.
• Extracted features and evaluated them using four conventional machine learning classifiers.
• Identified optimal channel-specific features for each emotional case to minimize data usage.

Main Results:

• Achieved high classification accuracies: 83-92% on DEAP/MAHNOB datasets and 78% on SEED dataset with minimal data.
• Optimal features were primarily located in the frontal region with diverse rhythmic characteristics.
• Individual differences were observed, with optimal features varying across subjects.

Conclusions:

• The proposed method significantly reduces complexity and data requirements for EEG-based emotion recognition.
• Insights gained advance the understanding of brain rhythms for diverse BCI applications.
• The approach facilitates the design of portable BCI devices requiring minimal electrodes.