The clutter Doppler frequency estimation method based on the auto-correlationfunction ry（1） is commonly employed in practical radar. The performance of this method degrades

however

in bimodal clutter einvironment due to the bias of the estimation. In this paper we presenta cumulant-based frequency estimation method under the assumption that the bimodal clutter consists of a Gaussian distributed clutter and a non-Gaussian distributed one. The method makes a fullutilization of all non-redundant sample fourth-order cumulants corresponding to the one-lag cumulant K4y（1）

a counterpart of ry（1） in cumulant domain. It is shown by asymptotic analysis and simulation experiments that the method is unbiassed and outperforms the conventional correlationbased method and the cumulant-based 1-D slice method with reduced estimation mean square error.