Snoring and apnea detection based on hybrid neural networks

Abstract

Snoring sound is an essential signal of obstructive sleep apnea (OSA). In order to detect snoring and apnea events in sleep audio recordings, a novel hybrid neural networks based snoring detection methods are evaluated in this study. The proposed method using linear predict coding (LPC) and Mel-Frequency Cepstral Coefficients (MFCC) features. The dataset included full-night audio recordings from 24 individuals who acknowledged having snoring habits with the label of polysomnography result. This method was demonstrated experimentally to be effective for snoring and apnea event detection. The performance of the proposed method was evaluated by classifying different events (snoring, Apnea and silence) from the sleep sound recordings and comparing the classification against ground truth. The proposed algorithm was able to achieve an accuracy of 90.65% for detecting snoring events, 90.99% for Apnea, and 90.30% for silence.