Classification of vibratory patterns of the upper airway during sleep

Upper airway (UA) narrowing and collapse during sleep results in obstructive sleep apnea (OSA). We hypothesize that vibratory patterns of snoring can distinguish simple snorers from those with OSA. Samples of breath sounds were collected from 7 snorers without OSA and 5 with OSA. Snoring pitch (F₀) contours were found using the robust algorithm for pitch tracking (RAPT). The OSA snoring contours showed fluctuating patterns as compared to the smoother patterns of simple snorers. This suggests that snoring reveals the underlying instabilities of UA tissue in OSA. Conditional random fields, a statistical sequence classifier, gave 75% accuracy in distinguishing the 2 groups.