Transitions in effective scaling behavior of accelerometric time series across sleep and wake

We study the effective scaling behavior of high-resolution accelerometric time series recorded at the wrists and hips of 100 subjects during sleep and wake. Using spectral analysis and detrended fluctuation analysis we find long-term correlated fluctuations with a spectral exponent ( noise). On short time scales, β is larger during wake () and smaller during sleep (). In addition, characteristic peaks at 0.2–0.3 Hz (due to respiration) and 4–10 Hz (probably due to physiological tremor) are observed in periods of weak activity. Because of these peaks, spectral analysis is superior in characterizing effective scaling during sleep, while detrending analysis performs well during wake. Our findings can be exploited to detect sleep-wake transitions.