Abstract: The method for optimizing light and sound programs for a falling asleep phase and for an awakening phase of a first user, in a system comprising, for the first user and other users, a bedside device, a bio parameter sensor, a smartphone, and additionally for all users a central server, with: /a1/ collecting, with regard to the first user, sleep data and sleep context data, said sleep data comprising at least light and sound program played for the falling asleep phase and for the awakening phase, bio parameters and sleep patterns sequence deduced therefrom, said sleep context data comprising at least previous daytime activity such as, sending this data to the central server, /a2/ repeat /a1/ for other users, /b1/ building a user-specific model of each user sleep behavior, /c/ comparing user-specific models to define groups of similar users, each group of users being allocated with a group meta-model with decision rules and preferred playlist of sound tracks, /d/ sending the group meta-model from the server to