A generalized normalization technique for signal detection in nonstationary correlated noise

A generalized normalization technique is presented for detecting random signals in a nonstationary, two-dimensional noise field. By comparing the covariance structure of the data under test with that of reference data, the test exploits information in both amplitude and phase of the data. As a result, the proposed scheme considerably outperforms the conventional normalization technique which is based only on the amplitude information. Theoretical performance has been established and its correctness has been confirmed through computer simulations