Single Channel EEG Based Sleep Stage Prediction Using CNN-CNN and CNN-LSTM

Sleep is one of the important things in human health and a critical aspect of physical and mental well-being. Adequate sleep is necessary for humans. The assessment of sleep stages through Polysomnography (PSG) is a widely used method to evaluate sleep quality. Sleep stage sequencing is critical for identifying sleep disorders, which is usually done by trained professionals by observing Electroencephalogram (EEG), Electrocardiogram (ECG), and Electromyogram (EMG) belonging to PSG data. This data is obtained by recording the brain waves from the scalp of the object. In this work, single-channel EEG data is analyzed to identify the sleep stages in 5 categories using supervised learning. With the fast growth in Artificial Intelligence (AI), it is now easy to design and develop Deep Learning (DL) techniques to automatically classify sleep stages. DL models offer high computational efficiency and have shown promising results in the field of bio-medical signal processing. In this work, considered well-known architectures CNN and LSTM and cascaded them as CNN-CNN & CNN-LSTM. The proposed cascading models exhibited good performance on PSG data taken from physiobank (Sleep EDFX) for sleep stage classification. The proposed model yields with accuracy of 89 % for CNN-CNN and 84 % for CNN-LSTM.