Sleep monitoring approach based on belief rule-based systems with pulse oxygen saturation and heart rate

This paper aims to propose a sleep and wake classification approach based on Belief Rule-Based (BRB) systems (called BRB classifier) to address the challenge of daily personal sleep monitoring at home. The BRB classifier infers sleep and wake states with two signals, pulse oxygen saturation and heart rate which can be obtained from wearable devices. The actual sleep datasets collected by Sleep Heart Health Study group are investigated to testify the efficiency of BRB classifier. In this case study, several experiments are conducted to determine the settings of BRB systems. By comparing with Linear Discriminate classifier and Neural Network classifier, the proposed BRB classifier has shown superior performance in sleep/wake classification with pulse oxygen saturation and heart rate.