Quantifying respiratory variation with force sensor measurements.

Measuring the variation of the respiratory rate makes it possible to analyze the structure of sleep. The variation is high when awake or in REM sleep, and decreases in deep sleep. With sleep apnea, the respiratory variation is disturbed. We present a novel method for extracting respiratory rate variation from indirect measurements of respiration. The method is particularly suitable for force sensor signals, because, in addition to the respiratory phenomenon, they typically contain also other disturbing features, which makes the accurate detection of the respiratory rate difficult. Respiratory variation is calculated by low-pass filtering a force sensor signal at different cut-off frequencies and, at every time instant, selecting one of them for the determination of respiration cycles. The method was validated with a single-night reference recording, which showed that the proposed method detects the respiratory variation accurately. Of the 3421 calculated respiration cycle lengths, 95.9% were closer than 0.5 seconds to the reference.