Method, system and computer program product are provided for estimating a circadian phase of a subject by: obtaining a sensed biological signal for the subject; and using, by one or more processors, adaptive frequency tracking to adaptively estimate the circadian phase of the subject from the sensed biological signal. Circadian phase estimation may be accelerated by providing a feedback loop for the adaptive frequency tracking, which utilizes, in part, a circadian phase model in automatically ascertaining a phase correction for the adaptive frequency tracking. The circadian phase estimation may be used in automatically constructing a light-based circadian rhythm model for the subject using a linear parameter-varying (LPV) formulation, and once constructed, the circadian rhythm model for the subject may be used to provide light-based circadian rhythm regulation.