Sleep Stage Classification Using Convolutional Neural Networks with Multichannel Physiological Signals

Abstract

Polysomnography (PSG) is a comprehensive recording technique that monitors physiological signals. Traditionally, collecting large amounts of PSG data has been challenging. The Sleep Heart Health Study (SHHS) dataset, recently made available, includes an extensive collection of polysomnography data with multiple physiological signals. This paper presents a novel approach to classifying the sleep stages in imbalanced and balanced datasets using convolutional neural networks. We then compare the classification performances of these two datasets. Additionally, we investigate the effect of interpolated and non-interpolated signals on classification performance.