Obstructive sleep apnea detection based on unsupervised feature learning and hidden markov model

摘要

Obstructive sleep apnea (OSA) is a sleep-related respiratory system disease which leads to increased risk of cardiovascular disease. Electrocardiogram (ECG) based method has been studied as a useful method in OSA detection. Previous studies were widely focused on feature engineering which had some disadvantages such as highly dependent on experts' priori knowledge. In this study, a method which was based on deep learning and Hidden Markov model (HMM) was proposed for OSA detection. In this method, sparse autoencoder was used for features learning and extraction, the support vector machine was used for features classify. Considering the temporal dependence of ECG signal, the HMM was adopted to improve the performance of classifier. Classification accuracy was achieved to 84.7% in per-segment detection. This result demonstrated that the method used in this study was reliable for OSA detection.