Signal detection in impulsive noise based on stable distributions

The previously developed symmetric stable model for impulsive noise is applied to the detection of known signals in additive noise. The structure and performance of the locally optimum detector are determined for stable noise and illustrations are presented to indicate behavior as a function of noise model parameters. The performances of selected linear and nonlinear suboptimum detectors, such as the clipper, hole puncher, or hard limiter, are also evaluated for stable noise and compared with that of the locally optimum detector using the finite sample approach. The results indicate that the locally optimum detector performs significantly better than the commonly used linear detector in the presence of additive stable noise.<<ETX>>