Automatic Sleeping Time Estimation and Mild TBI Detection Using Actigraphy Data

he sleep schedule and circadian phase irregularity are indicative of some health problems and diseases, e.g., narcolepsy, circadian disorder and concussion. Actigraphy has been widely used in the study of sleep and circadian rhythm. This paper presents a method for estimating the asleep/awake states based on the minute-by- minute actigraphy data measured by Actiwatch 2 from Philips Respironics. Compared with the scoring methods used in Philips Actiware, the estimated sleeping time from our method is more consistent with the sleep logs reported by the subjects. The circadian phase shift is estimated from actigraphy data using an adaptive notch filter algorithm. Concussion detection based on the sleep-related features calculated from the actigraphy data is finally discussed and detection accuracy is up to 90%, which implies that the concussion is closely related to sleep and circadian disorders.