Abstract-The estimation of the directions-of-arrival (DoAs) of multiple signals is a topic of great relevance in smart antenna synthesis and signal processing applications. In this paper, a memory-based method is proposed to compute the maximum likelihood (ML) DoA estimates. Such a conceptually-simple technique is based on the datasupported optimization (DSO) and the estimation of signal parameters via rotational invariance technique (ESP RIT ), but fully exploits a memory mechanism for improving the estimation accuracy especially when dealing with critical scenarios characterized by low signal-tonoise ratios (SN R) or/and small number of snapshots. Simulation results assess the potentialities and limitations of the proposed approach that favorably compares with state-of-the-art methods.