Apnea detection with microbend fiber-optic sensor

Abstract

As a common chronic disease, sleep apnea syndrome (SAS) seriously threatens patients’ health, so it is imperative to find an effective way of apnea detection. In this paper, we present a new method that explores of using the high sensitivity characteristic of the micro-bending effect of the gradient multimode fiber to detect human breathing movement, and thereby collecting the respiratory signals. With characterizing central apnea, 8 different features in time-domain and frequency-domain are extracted from the respiratory signals, which are used to train the classifiers. In the feature modeling, we present a peak calculation method based on moving average curve (MAC) to increase the accuracy of estimating the respiratory frequency and amplitude. In our experimental studies, the forward sequence selection method (SFS) is employed to combine these features for training SVM classifier, and our approach can reach an accuracy rate of 94.6% in apnea discrimination.