This paper deals with the problem of estimating the Direction-of-Arrivals (DOAs) of multiple wideband sources using an array of sensors. While a variety of estimation techniques have been proposed in the literature, the Maximum-Likelihood (ML) DOA estimator has been shown to have superior performance under many challenging environments. In this paper, we propose a novel implementation for ML DOA estimator based on the Cross-Entropy (CE) method. Simulation result shows the CE algorithm converges to the CRB in all scenarios within several iterations and the convergence speed is insensitive to the coherence of sources compared to the Alternating Projection (AP) method.