Sleep stage classification by body movement index and respiratory interval indices using multiple radar sensors

Disturbed sleep has become more common in recent years. To increase the quality of sleep, undergoing sleep observation has gained interest as an attempt to resolve possible problems. In this paper, we evaluate a non-restrictive and non-contact method for classifying real-time sleep stages and report on its potential applications. The proposed system measures body movements and respiratory signals of a sleeping person using a multiple 24-GHz microwave radar placed beneath the mattress. We determined a body-movement index to identify wake and sleep periods, and fluctuation indices of respiratory intervals to identify sleep stages. For identifying wake and sleep periods, the rate agreement between the body-movement index and the reference result using the R&K method was 83.5 ± 6.3%. One-minute standard deviations, one of the fluctuation indices of respiratory intervals, had a high degree of contribution and showed a significant difference across the three sleep stages (REM, LIGHT, and DEEP; p <; 0.001). Although the degree that the 5-min fractal dimension contributed-another fluctuation index-was not as high as expected, its difference between REM and DEEP sleep was significant (p <; 0.05). We applied a linear discriminant function to classify wake or sleep periods and to estimate the three sleep stages. The accuracy was 79.3% for classification and 71.9% for estimation. This is a novel system for measuring body movements and body-surface movements that are induced by respiration and for measuring high sensitivity pulse waves using multiple radar signals. This method simplifies measurement of sleep stages and may be employed at nursing care facilities or by the general public to increase sleep quality.