On the probability density of signaltonoise ratio in an improved detector
Abstract
We derive an approximate probability distribution for the SNR (signaltonoise ratio) of an improved adaptive detector in near rank1 Gaussian noise where the filter weights are computed using the principal eigenvectors of the estimated noise covariance matrix. The noise consists of two components, a strong rank1covariance interference component plus white noise. Computer simulation is used to verify the approximating SNR distribution and show that it is accurate even for small sample size and low interferencetonoise ratios (INR). We use this distribution to show the improvement possible when using filter weights based on just the principal eigenvectors rather than the full inverse of the estimated sample covariance matrix when the noise covariance is near rank 1. For example we compare the expected value of SNR for our improved method and the conventional adaptive detector based on the inverse of the estimated covariance matrix. We find that for 20 through 50 samples and an INR value of 10 dB, the expected value of SNR for our new method is better than the comparison method. Other statistics can also be obtained from the probability density. (Author).
 Publication:

NASA STI/Recon Technical Report N
 Pub Date:
 February 1985
 Bibcode:
 1985STIN...8528232K
 Keywords:

 Correlation Detection;
 Probability Distribution Functions;
 Random Noise;
 Signal To Noise Ratios;
 White Noise;
 Adaptation;
 Comparison;
 Computerized Simulation;
 Covariance;
 Eigenvectors;
 Probability Theory;
 Weighting Functions;
 Electronics and Electrical Engineering