Distributed Target Detection Using Samples Filtered with Normalized Conjugate Signal Steering Vector

This paper studied the problem of adaptive detection of distributed targets in colored noise with unknown covariance matrix (CM), for the case where limited noise-only (training) data are available to estimate this CM. We first filter the test and training data with the normalized conjugate signal steering vector which is matched to the target signal, to preserve the signal power while suppressing the noise power; second, we derive the generalized likelihood ratio test. The new detector has the desired constant false alarm rate feature against the noise CM; it needs less training data, has a lower computational complexity and performs better (more robust) for matched (mismatched) signals, when compared with its natural counterparts.