Auditory Image Model for the Characterisation of Obstructive Sleep Apnoea

Snoring is one of the major symptoms of Obstructive sleep apnea syndrome (OSAS) which is associated with serious health consequences. In recent years, speech-like analysis of snores has been performed towards an alterative to polysomnography (PSG) which is considered the gold standard in the diagnosis of OSAS. Human auditory perception has the ability to distinguish between different sounds. In this study, we propose a new method to extract snores from the sound recordings and characterize OSAS using auditory image model, which represents human auditory processing. Our proposed techniques resulted in a sensitivity of 98.4% and specificity of 94.06% for snore and non-snore classification and sensitivity of 85.00±26.87 at a specificity of 95.00±15.81 % for OSAS and non-OSAS classification. These results may indicate the feasibility of developing an OSAS population-screening tool based on automated analysis of non-contact snore.