Recognition of breathing route during snoring for simple monitoring of sleep apnea

This paper proposes a novel method to recognize breathing route during snoring (mouth, nose, and both) using the spectral properties of snoring sounds. Recent studies have reported that nasal closure determines an increase in the number of apneas during sleep. Thus, it is expected that if the seriousness of Obstructive Sleep Apnea Syndrome (OSAS) can roughly be estimated by a frequency of oral and oronasal snores at night, a simple device with only a microphone can automatically diagnose OSAS at bedside. In order to realize such device, it is necessary at first to recognize oral, nasal, and oronasal snores from the recorded sounds. Our proposed method can classify snoring sounds into nasal with 85.5%, oral with 86.1%, and oronasal with 70.7% respectively.