Musical Emotions Recognition Using Entropy Features and Channel Optimization Based on EEG

Area of Science:

• Neuroscience
• Cognitive Science
• Signal Processing

Background:

• Current emotion recognition research often uses artificial stimuli, failing to capture complex, dynamic emotional responses.
• Music, with its inherent dynamic emotional qualities, offers a more naturalistic stimulus for studying brain responses to emotion.

Purpose of the Study:

• To investigate the brain's dynamic response to complex emotions evoked by long-term musical stimuli.
• To develop and validate a method for emotion recognition using electroencephalogram (EEG) signals and non-linear feature extraction.
• To identify a common set of EEG channels that represent universal emotional features across subjects.

Main Methods:

• Utilized three long-term musical pieces with diverse emotional content as stimuli.
• Applied Approximate Entropy (ApEn) and Sample Entropy (SampEn) to extract non-linear features from EEG signals.
• Employed K-Nearest Neighbor (KNN) for emotion classification and Particle Swarm Optimization (PSO) for optimal channel selection.
• Proposed a supervised feature dimensionality reduction method to identify common channels across subjects.

Main Results:

• Emotion recognition accuracies exceeded 80% for simpler emotional categories and approximately 70% for more complex ones, using optimal channel sets.
• Emotion recognition using a common channel set achieved accuracies about 10% lower than optimal sets but comparable to using all channels.
• The frontal lobe showed a higher concentration of common channels, indicating its significant role in emotional processing.
• Topographic maps revealed distinct entropy patterns across brain regions for different emotions.

Conclusions:

• Long-term musical stimuli effectively evoke complex emotions and their dynamic brain responses.
• A common channel set derived from optimal subject-specific channels can achieve efficient emotion recognition while reducing dimensionality.
• The frontal lobe plays a crucial role in processing music-induced emotions.
• This approach provides a foundation for developing more sophisticated brain-computer interfaces for emotion detection.