Higher-order-statistics-based sharpness evaluation of a generalized Gaussian PDF model in impulsive noisy environments

The aim of the present work is to provide a simple and reliable mathematical method in order to evaluate the sharpness parameter of a generalized Gaussian PDF model, as a function of a higher order statistic, namely the normalized kurtosis. Well-known detection and estimation algorithms in non-Gaussian noise are based on the accurate modelling of the distribution function of the noise, made on the basis of the higher-order-statistics (HOS) parameters estimated from the incoming data. An application of the proposed evaluation to multilevel signal estimation in non-Gaussian impulsive noise is then considered.