Abstract
This paper presents a novel technique to detect obstructive sleep apnea episodes in infrared video recordings. The major advantage of the proposed approach is that it does not require any manual adjustments and does not depend on the patient pose. Our contribution is threefold. First, the detection of the breathing movement is based on a robust estimation of the optical flow. Second, we resort to the use of the summed motion magnitudes as an important feature allowing to distinguish between normal breathing episodes and obstructive sleep apnea episodes. Third, the motion magnitudes during apnea events are considered as having atypical values relatively to those obtained during a normal breathing and are detected thanks to the use of a statistical test for outliers detection. The experimental evaluation on real videos show high performances of the proposed technique.