Abstract: Sleep apnea syndrome is a common and potentially harmful sleep disorder, and the classification and detection of sleep apnea can provide an important basis for the diagnosis of the disease. Due to their non-contact nature, video-based sleep monitoring systems are universally applicable for disease screening, among which thermal imaging cameras, with strong privacy protection, have attracted wide attention in recent years. In this paper, we propose a novel sleep apnea detection and classification method using thermal imaging. By obtaining the temporal information of thoracic and abdominal movement, a two-dimensional complex feature space mapping central and obstructive sleep apnea under different physiological mechanisms is constructed. Based on their statistical properties, the respiratory effort intensity feature and the respiratory effort asynchrony feature are proposed to achieve the classification and detection of two types of sleep apnea. Experimental results show that the accuracy of detecting both types of sleep apnea exceeds 97.0%. This work effectively overcomes the problem of difficulty in extracting valid information caused by observation noise and redundant information in videos, and is expected to assist in the actual screening and diagnosis of sleep disorders.

Key words: sleep apnea classification, thoracic and abdominal movement, thermal imaging, respiratory effort intensity, respiratory effort asynchrony