Detecting nonlinearity in prediction residuals of snoring sounds
Tsuyoshi Mikami · 2009 ICCAS-SICE · 2009
This paper focuses on the nonlinear properties of snoring sounds for the purpose of obstructive sleep apnea diagnosis. Snoring sounds are convolutional sounds caused by wheezing of airway obstruction and oscillation of soft palate. Namely, it should be natural that the snoring sounds are generated from a nonlinear dynamics, but the nonlinear properties of them have not yet been studied so far. In this paper, the nonlinearity is defined as the predictability using a linear AR prediction model, and the prediction residuals are analyzed by portmanteau test.