Bispectral Analysis of Snore Signals for Obstructive Sleep Apnea Detection

Obstructive sleep apnea (OSA) is an insidious condition of recurring upper airway closure during sleep. Apart from polysomnography, many researchers tried to explore alternative methods to detect OSA. However, not much work has been done to address the non-Gaussian and nonlinear behavior of the snore signals, which the power spectrum may not adequately account for. Therefore, this paper presents the use of bispectral analysis of snore signals for OSA detection. The raw snore signals were denoised using a modified level-wavelet-dependent thresholding scheme under an undecimated wavelet environment. Subsequently, nonlinear properties in the noise-suppressed snore signals were extracted to discriminate between apneic and benign snores. Results show that apneic snores exhibit higher degree of phase coupling phenomena than benign snores. This preliminary study suggests that the bispectral analysis of snore signals might be useful to distinguish apneic patients from benign patients.