Spectrum based covariance estimation for nongaussian multipulse detectors

Multipulse detection under nongaussian clutter is addressed. GLRT based detectors have already been developed for this problem. These detectors require the correlation estimation of the local environment for practical use. Some ML based estimators have been developed for this purpose. Unfortunately the exact ML estimate does not exist for the compound gaussian clutter model considered in this study. Only approximations exist. In this work this problem is handled with a different approach. Correlation estimation problem is reformulated as a local doppler spectrum estimation problem. After the Doppler spectrum estimated, correlation estimate is obtained from spectrum-correlation relation. Proposed estimators are compared with the ML based estimators and it is shown that the performances of the proposed estimators are equivalent or superior to the ML based estimators in many conditions.