We present an identification algorithm for determining the variables characterizing a circadian rhythm. This method is based on a recursive least square algorithm combined with an appropriate observer. It has the advantage of determining all the parameters of a time-varying circadian rhythm without requiring the knowledge of the period. A comparison study with respect to the Lomb & Scargle method has been carried out to show the effectiveness of the proposed method. The presented algorithm has been applied to data collected during a laboratory experiment.