Generalized canonical correlation for passive multistatic radar detection

In this paper, we consider the problem of target detection in passive multistatic radar. In passive radar, we make use of illuminators of opportunity. As the illuminators are not under our direct control, the illuminating signal itself is unknown. We propose a signal model which reflects this. In deriving a maximum-likelihood estimator for the unknown parameters, including the illumination, we find that the maximum value of the likelihood is a monotonic function of the largest eigenvalue of the Gram matrix of the received signals. The generalised likelihood ratio test turns out to be equivalent to comparison of the largest eigenvalue against a threshold, so we propose its use as a target detection statistic. The proposed detector is similar to generalised canonical correlation in multivariate statistics. The benefit of using this statistic over others such as generalised variance is demonstrated through numerical simulations in the context of passive radar using DVB-T signals.