A near-optimal receiver for detection in α-stable distributed noise

Abstract

Until recently, work on detection theory considered only detection of signals in Gaussian noise. However, it is very well-known that various types of noise such as atmospheric noise and underwater acoustic noise which are the governing types of noise in radar and sonar applications show an impulsive character. Recently, there has been widespread interest in the /spl alpha/-stable distributions for modelling impulsive noise. Unfortunately, it has not been possible to design receivers for optimal detection in /spl alpha/-stable noise since /spl alpha/-stable distributions in general cannot be represented with compact analytical expressions. Recently, it has been suggested that this drawback can be removed by representing the /spl alpha/-stable PDF with finite mixtures of Gaussians. We suggest near-optimal receivers for detection in /spl alpha/-stable noise based on this new representation. The simulation results demonstrate that the new detector has significantly better performance than the previously suggested ones and has a very close performance to the optimal detector.