Assessment of the single data set detection algorithms under template mismatch

Abstract

The detection of signals with known templates embedded in zero-mean coloured Gaussian interference is relevant to many fields such as radar, sonar, seismology and biomedicine to name a few. Traditional detection algorithms, such as the GLRT and AMF, require a training data set. Recently, single data set (SDS) algorithms, namely the GMLED and MLED, have been proposed to deal with the case where training data may not be available. In this paper, we examine the performance of these algorithms under template (or steering vector) mismatch. We identify three types of mismatch, namely the spatial steering vector mismatch, temporal steering vector mismatch and mismatch in both steering vectors. In each mismath case we derive the expected signal to noise ratio loss with respect to the corresponding matched case. Simulation results are given which show that the SDS algorithms are more sensitive to mismatch mainly due to the interaction between the signal and subspaces estimation. However, this increased sensitivity to mismatch is closely related to the ability to resolve close signals. Therefore, the SDS algorithms exhibit higher resolution