In the multi-object localization, most of the existing methods are based on accuracy time - delay estimation. In this paper a maximum - likelihood multiple sources localization estimation method is presented. The ML method can be used in estimating multiple source locations, while such case is rather difficult for the time - delay method. In the single source case, the ML estimator is shown to be equivalent to maximizing the sum of the weighed cross-correlations between time shifted sensor data. Numerical examples using simulating data demonstrate the effectiveness of the new method.