Nonlinear properties of snoring sounds

In this paper, the Gaussianity and linearity of the snoring sound (SS) segments extracted from respiratory sounds are discussed. The respiratory sound signals were recorded from 30 individuals by two microphones simultaneously with full-night Polysomnography (PSG) during sleep. The first microphone was placed over the trachea (the tracheal microphone), and the other one was a free standing microphone (the ambient microphone). The SS segments were identified automatically from the respiratory sounds. The bispectrum and bicoherence of each SS segment were estimated. These measures were used to construct a statistic to test for the Gaussianity and the linearity of each SS segment. The result showed that all SS segments recorded by the tracheal microphone were non-Gaussian, while their linearity was variable over time. It was also observed that the number of nonlinear SS segments were higher in the sounds recorded by tracheal microphone than in those recorded by the ambient microphone.