Maximum likelihood array processing: the semi-blind case

Protocols for digital wireless communication systems generally contain known bit sequences that are transmitted periodically for the purposes of training and/or synchronization. We study the performance improvement that is obtained in antenna array processing when knowledge of the signal waveforms is exploited. Our model is formulated such that the signals from multiple sources are known for a subset of the data samples and unknown for the remainder of the samples. Maximum likelihood (ML) estimates of the signal parameters are derived for this model, and the Cramer-Rao bound (CRB) on the accuracy of source location estimates is presented. Numerical evaluations of the CRB are included to illustrate the performance when partial training data is available in one or more of the sources.