Abstract
Individuals who suffer from different sleep breathing disorders suffer from a wide range of serious health problems. Unfortunately, the rate of diagnosis is very low, and the existing breathing monitoring techniques are expensive, uncomfortable and time- and labor-intensive. The gold standard PSG is invasive, costly, technically complex and time-consuming. Toward developing a non-contact sleep breathing monitoring system, this study presents a motion-based computer vision approach that aims to detect breathing movements of the sleeping patient from infrared videos and map them into a waveform. The proposed waveform illustrates that each type of breathing difficulty has a specific pattern and hence can be easily distinguished. This facilitates identifying only suspicious periods during which physiological signals will be scored, instead of analyzing the whole signals of 8h of sleep.