Adaptive Reduction of Additive Noise From Sleep Breathing Sounds Tech Report: CSLU-2012-001

Sleep-disordered breathing is believed to be a widespread, under-diagnosed condition associated with many detrimental health problems [1, 2]. Young, et al. describe the total burden of sleep-disordered breathing on the health system and society as “staggering” [3]. The current gold standard for diagnosis of sleepdisordered breathing is a sleep study, or polysomnography (PSG). This overnight procedure takes place in a sleep laboratory and is obtrusive, typically recording twelve or more biological processes (e. g., electroencephalogram, electrooculogram, electromyogram, blood oxygen saturation, nasal airflow) while requiring 22–40 wires to be attached to the patient. Scoring of study results is also time-consuming and expensive, as an entire night-long study must be manually assessed by a human expert, then reviewed by a clinician to determine a diagnosis. Moreover, studies show that patients sleep differently at a hospital or clinic than at home [4]. The complex clinical nature and high cost of PSG make the procedure ill-suited for mass screening of the population. Consequently, there is a tremendous need for an alternative method to screen for sleepdisordered breathing. Our current work investigates an acoustics-based system for tracking breathing during sleep in a patient’s home sleep environment. This system detects long pauses in the breathing cycle and episodes of intense, frequent snoring. Acoustic recordings made in a patient’s home sleep environment are highly susceptible to additive noise. Background noise present during signal collection can lower the performance of a sound classification system. Air conditioners and furnaces are typical sources of this type of noise in a home environment. During the course of a single night, an air conditioner or furnace may turn on and off many times, obscuring sleep breathing sounds. This paper presents a method for adaptive reduction of additive noise from sleep breathing sounds to increase breath and snore classification accuracy.