Adaptive Alarm Clock Using Movement Detection to Differentiate Sleep Phases

Waking up can be very troubling. In many cases, being awoken by the alarm clock during a period of deep sleep causes a person to not feel fully rested after waking up. This paper describes how an adaptive alarm clock was developed addressing this problem. The alarm clock predetermines in what state the sleeper will be at the instant of supposed alarm-firing, and adjusts that instant to a more favourable one. Preferably for the sleeper to be awakened during a period of light sleep. For this the system gathers information about the sleeper during the night via a webcam. The videofeed is processed in real-time on a Nvidia videocard using CUDA, determining the amount of movement by the slumberer at that instant, which is a direct indication of which sleep phase he or she is currently experiencing. This information can be analyzed to predict future phases, to adjust the instant of alarm-firing. The possibility to use any webcam and a standard Nvidia GPU, makes this a solution available to a wide audience.