Independent sample mean squared error for adaptive detection statistics

This paper introduces the independent sample mean square error (ISMSE) as an important factor in computing the adaptive matched filter (AMF) constant false-alarm rate (CFAR) detection statistic. It has been shown that the ISMSE is a cross-validation metric that is useful for determining the optimum rank of the multistage Wiener filter (MWF) subspace. This is because it gives a more accurate estimate of the true ensemble mean squared error (MSE) provided by a Wiener filter, as compared to the standard least squares sample mean squared error (SMSE). The innovation described herein directly exploits this improved MSE estimator within the AMF calculation. Together with the reduced rank MWF solution, this results in improved detection performance in low sample support environments. Receiver operating characteristic (ROC) curves are generated using Monte Carlo simulations and used to assess performance against the full rank approach