Abstract: Class B noise model is statistical-physical model for atmospheric noise. This paper proposes a nonlinear regression-type algorithm to estimate the parameters of Class B model. The algorithm based on characteristic function is derived from nonlinear regression model, which has low iterations. An initial estimators are also designed to accelerate the convergence of algorithm, and a special series are utilized to calculate the log characteristic function to solve the issue of many zero points of characteristic function. The result shows that new method has high precisions and low iterations, which can be applied excellently to practice.