A Correlator with Preposed Whitening and Gaussianizing Modules of Signal in Colored Non-Gaussian Interference Background

A correlator with preposed whitening and Gaussianizing modules (CPWG) is proposed, which is an asymptotically optimal detection of signal in colored non-Gaussian interference background. Gaussian mixture autoregressive model (GMAR) is adopted to fit the probability density and power-spectrum density of the colored non-Gaussian processes. A coupled estimation algorithm, LSEM, is developed, which can estimate both probability density function parameter and power-spectrum density function parameter of GMAR at the same time. Based on GMAR parameter estimation, prewhitening and Gaussianizing modules are set up. The observed processes pass through these two modules firstly, and then being send into the correlator. Processing results of simulation and lake trial data are demonstrated. A conclusion can be reached: CPWG’s performance is much better than the classic match filter, and even better than the generalized match filter which has a prewhitening module. Lying on the non-Gaussianity intensity, to get the same level of output signal-noise-ratio, the input signal-noise-ratio which CPWG requires is lower than generalized match filter from several to several decades dB.