Advanced Sleep Stage Classification Through a Convolutional Recurrent Attention-Based Neural Network Model

Sleep stage classification plays a critical role in diagnosing sleep disorders, which are linked to cognitive and physical health issues. Polysomnography (PSG) is commonly used for this purpose, but manual classification is time-consuming and prone to error. This paper proposes a hybrid deep learning model combining Convolutional neural networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM) networks with attention mechanisms to classify PSG signals from single electroencephalogram (EEG) channels into five sleep stages. The model captures both spatial and temporal features from the EEG data, addressing class imbalance through random under-sampling. Experimental results on the Sleep-EDF dataset demonstrate the effectiveness of this approach, with promising accuracy and F1-score metrics, outperforming traditional methods and offering improved classification robustness.