Secondary data partitioning via maximum likelihood estimation: extensions

Adaptive beamforming based on sample matrix inversion requires the availability of a secondary data set for "training" (i.e., estimating the sample covariance matrix). For each snapshot processed, near-optimal detection performance can be achieved as long as the training data is independent, identically distributed, and free of target-like signals. Oftentimes, however, the available set of secondary data is contaminated by an unknown heterogeneous mixture of clutter, jamming, electromagnetic interference, and strong targets. Previously, the author and his colleagues applied maximum likelihood estimation (MLE) to pre-screen the available secondary data, identifying subsets having similar clutter and jamming statistics, and using these subsets for adaptive beamformer training. This paper extends the basic MLE approach to cases where the secondary data may contain electromagnetic interference and/or strong targets.