In this paper, we discuss properties of global convergence in DOA finding based on Maximum Likelihood estimator using the stochastic signal model (abbreviated as SML). An adequate solution of SML is hardly obtained when using a global search method. This paper presents that the SML criterion does not have the uniqueness of the solution inherently. That is, many global minimum exist. Furthermore, it follows from the property of the SML criterion that in order to obtain an adequate solution of DOA, the use of a local search together with the initial value obtained by the deterministic ML is effective.
展开▼