Noncontact Sleep Stage Classification Based on Multi-sensor Feature Level Fusion

Abstract

This paper presents a new method to realize noncontact sleep stage classification which is based on multi-sensor feature level fusion. The method has be implemented by processing vital signs related to sleep and extracting features from continues wave (CW) Doppler radar sensor and audio sensor. And a novel feature level fusion model is proposed and trained based on machine learning algorithm. Assisting with Polysomnography (PSG) standard sleep stages, the feature level fusion model has achieved high accuracy in noncontact sleep stage classification.