Obstructive Sleep Apnea Analysis Based on Heart Rate Variability

Abstract

In this project we compared features in time segments of the heart rate to distinguish obstructive sleep apnea (OSA) from ‘normal’ sleep segments. The time duration of the segments is 5 minutes and features include estimates of the autocorrelation function (ACF), partial autocorrelation function (PACF), and power spectral density (PSD). In the OSA segments PSD estimator did not resolve the apnea breath/cessation cycles of the ensemble average, instead, shows a higher PSD in the apnea frequency band range (0.01-0.04 Hz). The ACF decays rapidly in a ‘normal’ sleep segments, while in the apnea segments the cycles were shown as a decaying sinusoid. The presence of only two major peaks in the PACF at 1 and 2 seconds lags indicates both apnea and ‘normal’ sleep processes have equal memory.