Development of sleep-wake estimation algorithm using the wrist acceleration sensor

We developed the algorithm to estimate sleep-wake during sleep period using a 3-axis acceleration wrist sensor for the purpose of self-management of sleep status in everyday life. The 3-axis acceleration data were transformed into combined acceleration time series, from which sleep-wake was estimated. The algorithm discriminated sleep from waking time with 91.8[87.4-95.3]% sensitivity and 55.5[40.2-72.5]% specificity. It estimated sleep efficiency with a correlation coefficient of 0.52(p<;0.01, n=228) with that determined by polysomnography.