In this paper a recursive technique, based on the subspace state space identification methods, is presented for identification of time-varying systems. The main idea was to develop an iterative algorithm with most of the advantages of this kind of methods in order to deal with real-time applications and minimize the computational burden. As a subspace-based state space system identification technique, it has two main steps: first, a state vector sequence is estimated, using numerical linear algebra tools, then the state space model is obtained from a simple least squares model.