Optimal multimodal feature combination and classifier selection for music-based EEG signal analysis

Area of Science:

• Neuroscience
• Cognitive Science
• Signal Processing

Background:

• Music perception is crucial for cognitive and emotional processing.
• Understanding neural dynamics of music perception aids in neuroscientific investigation.

Purpose of the Study:

• Examine neural dynamics in music perception.
• Identify optimal methods for mental state classification using electroencephalography (EEG) signals.

Main Methods:

• Utilized Indian classical music (ICM) to evoke distinct cognitive states.
• Extracted diverse EEG features (statistical, covariance, wavelet, fractal, entropy).
• Employed machine learning classifiers (SVM, KNN, perceptron, tree-based) optimized via Bayesian Optimization.

Main Results:

• Classification performance varied significantly with feature and classifier choice.
• Covariance features with KNN yielded high accuracy.
• Combined features and KNN achieved up to 98.65% accuracy, indicating shared and distinct neural patterns.

Conclusions:

• Music perception involves complex neural coordination across brain regions.
• Multimodal feature combinations enhance classification performance for mental states.