Sleep Stage Classification by Heart Rate Variability, Body Movement Index and Respiratory Interval Indices Using Non-contact Doppler Radars

Abstract

We proposed a technique of variable window-width of Moving-Average Method to extract only the pulse wave signals from the output signals of radars installed underneath the mattress, which enabled to calculate the heartbeat intervals.We applied a canonical discriminant analysis to classify the four sleep stages using the heart rate variability (HRV) derived from 10 students. LF/HF had a high degree of contribution to discriminate between LIGHT and DEEP sleep, and it was showed that the sleep cycle could be estimated (Figure).For the further improvement, we attempted to introduce a new decision tree analysis using not only HRV but also body movement index and respiratory interval indices.