We investigate the potential of k-nearest neighbor (KNN) based decision algorithms to detect a coherent signal in presence of non-Gaussian clutter, modeled in terms of a K-distributed spherically-invariant random vector (SIRV), plus thermal noise. The decision rule is fed by commonly used statistics, i.e., modified adaptive coherence estimator (ACE) and Kelly's statistics. The performance assessment shows that KNN based detectors can achieve intermediate performance between the modified ACE and Kelly's detectors for low signal-to-clutter ratio (SCR) values, and close to the latter for higher SCR values. A sensitivity analysis to possible mismatches of the clutter covariance matrix and/or the shape parameter of the K-distribution is also performed.
Radar detection in K-distributed clutter plus thermal noise based on KNN methods
Coluccia A.;Ricci G.
2019-01-01
Abstract
We investigate the potential of k-nearest neighbor (KNN) based decision algorithms to detect a coherent signal in presence of non-Gaussian clutter, modeled in terms of a K-distributed spherically-invariant random vector (SIRV), plus thermal noise. The decision rule is fed by commonly used statistics, i.e., modified adaptive coherence estimator (ACE) and Kelly's statistics. The performance assessment shows that KNN based detectors can achieve intermediate performance between the modified ACE and Kelly's detectors for low signal-to-clutter ratio (SCR) values, and close to the latter for higher SCR values. A sensitivity analysis to possible mismatches of the clutter covariance matrix and/or the shape parameter of the K-distribution is also performed.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
