11.07.2015 Views

Space/time/frequency methods in adaptive radar - New Jersey ...

Space/time/frequency methods in adaptive radar - New Jersey ...

Space/time/frequency methods in adaptive radar - New Jersey ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

67where k is a ga<strong>in</strong> constant and R is the colored noise true covariance matrix. TheWiener filter can be <strong>in</strong>terpreted as a cascade of a whiten<strong>in</strong>g filter for the <strong>in</strong>terference,followed by a matched filter. This solution requires the knowledge of thetrue covariance matrix R.3.2.3 Sample Matrix InversionIn practice, the true covariance matrix of the <strong>in</strong>terference R is unknown a priori andneeds to be estimated from a limited set of the secondary data (noise and <strong>in</strong>terference)as given <strong>in</strong> Equation 3.8.Substitut<strong>in</strong>g R for the true covariance matrix, R, <strong>in</strong> Equation 3.18 yields thefollow<strong>in</strong>g solutionThe covariance matrix used for this solution has been estimated from a f<strong>in</strong>ite numberof secondary samples. Therefore, this solution for the weights is not optimal.The conditioned signal-to-colored noise ratio (CSNR) is def<strong>in</strong>ed as the ratio ofthe actual signal-to-colored noise ratio (SNR) to the optimal SNR. The optimal SNRresults when the covariance is known. Colored noise refers to the white noise and<strong>in</strong>terference. If K 2N snapshots are used to estimate R <strong>in</strong> Equation 3.8, Reedet al.[15] showed that this solution <strong>in</strong> Equation 3.19 achieves a CSNR at the outputof the array with a mean of 0.5. A 3 dB loss with respect to the optimal SNR is aCSNR with a mean of 0.5. The CSNR is a random variable bounded between 0 and1 because the covariance matrix is estimated us<strong>in</strong>g Equation 3.8 and the result<strong>in</strong>gsignal-to-noise ratio is a random variable.3.2.4 Generalized Sidelobe Canceller (GSC)The Generalized sidelobe canceller (GSC) is a general form of a l<strong>in</strong>early constra<strong>in</strong>ed<strong>adaptive</strong> beamform<strong>in</strong>g algorithm[48]. It adapts to m<strong>in</strong>imize mean square error (MSE)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!