Time Series Analysis of Wrist Accelerometry and Cardiorespiratory Signals During Sleep

Résumé du livre

This cumulative dissertation deals with the analysis of cardiorespiratory biosignals in combination with high resolution accelerometry. In the first part, we focused on biosignals hidden in acceleration measurements. For wrist accelerometer data we have developed algorithms to extract the pulse wave propagating through the wrist, and also detect respiration movements during sleep. We found that pulse waves can be better reconstructed than respiration signals. Comparing sleep stages we found best reconstruction during deep sleep, while during wakefulness the reconstruction quality drops due to movement artifacts. The second part investigates the scaling behavior of cardiorespiratory signals. We found that heart and pulse rates are characterized by sex- and age-dependent short-term (6-16 s) fluctuations, while their long-term (50-200 s) fluctuations exhibit a clear sleep stage dependence. In contrast, pulse transit times do not show differences between short- and long-term scaling behavior.