Knowledge-aided covariance estimate via geometric mean for adaptive detection

Abstract This paper deals with the covariance estimation and its application in radar signal processing when the number of the secondary data is limited. We model the covariance estimation as a color loading version and use three different geometric means to derive three kinds of knowledge-aided (KA) covariance estimators, namely, KA Euclid (KA-E) metric estimator, KA Power-Euclid (KA-PE) metric estimator, and KA Log-Euclid (KA-LogE) metric estimator. Experimental results on simulation and measured data demonstrate that the proposed estimators achieve better performance than their natural competitors and keep good constant false alarm ratio (CFAR) property. Among the three proposed estimators, the KA-E estimator has the best performance.