Decision based non-linear filtering using interquartile range estimator for Gaussian signals

Abstract

Decision based non-linear filtering is widely used for the removal of impulsive noise. Various robust statistical estimators of scale are in use for determining the threshold of the filtering process. Real-time filtering requires this estimation to be computationally efficient and realizable within the system constraints. This paper proposes the use of Interquartile range (IQR) for filtering impulsive noise from the signals possessing Gaussian distribution. The efficiency of filtering using IQR has been described through simulation for varying levels of impulsive noise. An iterative threshold for IQR based filtering has been introduced to improve the filtering efficiency. A real-time technique for computing IQR on digital hardware has been introduced. The proposed technique has applications in the areas of passive microwave radiometry, digital communications and radio astronomy.