Classification of sleep stages using multi-wavelet time frequency entropy and LDA.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Automatic sleep stage scoring relies on feature extraction and classification from polysomnographic recordings, primarily electroencephalograph (EEG) signals.
• EEG signals are non-stationary, posing challenges for accurate wave detection and analysis.

Purpose of the Study:

• To present a new technique for automatic sleep stage scoring.
• To utilize continuous wavelet transform (CWT) with multiple mother wavelets and linear discriminant analysis (LDA) for enhanced EEG signal analysis.

Main Methods:

• Feature extraction was performed using CWT time-frequency entropy with three distinct mother wavelets.
• Classification of sleep stages was conducted using linear discriminant analysis (LDA).
• The proposed method was evaluated on 32 datasets from the MIT-BIH database.

Main Results:

• Successful extraction of EEG signal features using CWT time-frequency entropy with three mother wavelets.
• The proposed method demonstrated superior performance compared to CWT-based classification using a single mother wavelet.
• Achieved an accuracy of 0.84 and a kappa coefficient of 0.78.

Conclusions:

• Wavelet time-frequency entropy is an effective tool for feature extraction from non-stationary EEG signals.
• Utilizing multiple wavelets for feature extraction enhances the accuracy of sleep stage classification compared to single-wavelet approaches.