부산대학교 도서관

Obstructive sleep apnea (OSA) is a sleep disorder that causes abnormal and periodic breath interruptions during sleep. In this article, we present a novel one-dimensional (1-D) phase descriptor (PD) based methodology for single-lead electrocardiograph (ECG) based OSA detection. Specifically, we introduce two 1-D local PDs, referred to as 1-D local phase patterns and 1-D local phase quantization, and study their discriminative ability for OSA detection. Essentially, these PDs are computed using phase information extracted from the ECG signal. In our approach, PDs extracted from the Gabor filter phase responses are used to form the feature vector, which is utilized by a least-squares support vector machine (LS-SVM) for classification. Experimental evaluations on Physionet single-lead ECG dataset suggest that the proposed approach achieves state-of-the-art performance for OSA detection. Specifically, the proposed OSA detection approach achieves an average classification accuracy of 93.31% on ECG signals of 1-min duration, thereby providing a considerable improvement over the existing methods.