The statistical characteristics of some non-Gaussian noises were analyzed.On this ground the moments estimations were used to deduce the models transformation algorithms of these noises.The inertialess nonlinear transformer was employed to research the maximum likelihood detections of deterministic signals;and the recurrence was adopted to study the maximum likelihood estimations of cosine signals parameters in bimodal noises;on these grounds the validities of models transformation algorithms were verified by ways of analyzing the performance of signals detections and parameters estimations.The simulation results indicated that it was exact to regard the bi-peak noises as bimodal noises.Thus it was demonstrated that the thought of bimodal noise is practicable.