A Comparison between Two Parameter Estimation Methods for Non-Gaussian Noise Models

Atmospheric noise often often suffers from the impulsive noise, which seriously deviates from traditional Gaussian noise. When the signal processor is still developed with Gaussian noise, the performance would be seriously degraded. Before designing signal processors, the distribution of non-Gaussian noise often including many unknown parameters should be first estimated. Class A model is an important distribution for impulsive noise. However, this model mainly consists of three key parameters. In order to estimate three parameters of Class A model, parameter estimators should be designed. Up till now, the method based on characteristic function and that based on fast Fourier transformation (FFT) are exploited. In this paper, both parameter estimation methods are compared in detail. The processing results show that both methods are characterized by a small number of iterations. Besides, the performance of FFT method is superior to that of characteristic function method.