Testing the system non-linearity in snoring sound via neural networks

Obstructive sleep apnea (OSA) is a serious disease caused by the collapse of upper airways during sleep. OSA is almost always accompanied by snoring. While snoring is not currently used in the clinical diagnosis of OSA, there have been intense efforts recently to model snoring for that purpose. Conventional approach is to treat snores as the outcome of a linear process and apply techniques such as linear prediction coding (LPC). However, the snores are likely to have diagnostically relevant non-linearities that cannot be captured by linear techniques. In this paper, we investigate the non-linearity of snores and develop a novel measure, as a possible characterisation index. The method is based on artificial neural networks (NN). The developed method was tested on a database of 27 subjects (5568 snoring episodes), categorised into two groups based on their respiratory disturbance index (RDI)