Amplitude-Time Dual-View Fused EEG Temporal Feature Learning for Automatic Sleep Staging

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Electroencephalogram (EEG) signal recognition is crucial for automatic sleep staging systems.
• Challenges in EEG analysis include nonstationary characteristics and individual differences, hindering effective feature extraction.
• Developing robust EEG feature learning methods is essential for practical applications.

Purpose of the Study:

• To propose a novel temporal feature learning method for automatic sleep staging using EEG signals.
• To investigate the feature extraction capabilities of convolutional neural networks (CNNs) for EEG data.
• To enhance sleep staging performance by fusing dual-view EEG signal features.

Main Methods:

• Constructed two representation signals from raw EEG: amplitude and time views.
• Utilized 1-D CNNs to extract amplitude-time signal features reflecting sleep stage transitions.
• Employed a hybrid dilation convolution module to capture long-term temporal dependencies in EEG signals.
• Applied attention-based feature fusion to integrate dual-view features for improved performance.

Main Results:

• The proposed method demonstrated superior sleep staging performance on standard datasets.
• Experimental results validated the effectiveness of the amplitude-time dual-view fusion approach.
• The method showed potential for application in EEG-based automatic sleep staging systems.

Conclusions:

• The novel temporal feature learning method based on amplitude-time dual-view fusion effectively addresses challenges in EEG signal analysis for sleep staging.
• The approach enhances the accuracy and reliability of automatic sleep staging systems.
• This work contributes to the advancement of practical EEG-based diagnostic tools for sleep disorders.