A Computer Vision Approach to Analyse Sleep in Infrared Video

Sleep apnea is a respiratory disorder characterized by repetitive interruptions of breathing during sleep. About 10% of the population have sleep apnea. Depending on the presence of breathing effort, respiratory events during sleep are divided into obstructive and central events. The severity of sleep apnea is often associated with sleeping position. Choosing the appropriate treatment relies on identifying types of apneas and detecting sleep positions during apneas. Available diagnosis methods are expensive, need a trained technician to set up, and require attachment of sensors to the user, rendering them inconvenient during sleep. To address these problems, this thesis proposes non-contact computer vision-based algorithms that automatically analyze and track movement on infrared videos to detect sleep position and to differentiate obstructive from central apneas. These algorithms can be used to develop an affordable and easy-to-use home assessment technology that could help physicians choose suitable treatments.M.A.S