The application of adaptive multichannel spectral estimation for broadband array processing
An alternative means of reducing the observation time required to make high-resolution broadband spatial spectral estimates is presented which precludes the need for preliminary periodogram cross spectral density matrix (CSDM) estimates. The methods described are based on low order multichannel minimum variance (MCMV), and multichannel autoregressive (MCAR) spectral estimators. The low-order nature of the array output process is shown by examining the approximate rank of a Hankel matrix defined from the array output assuming low order autoregressive moving average models of source spectra. Results based on simulation and actual towed array data which indicate MCMV and MCAR offer a means of observation time threshold reduction and comparable performance with previously proposed methods based on preliminary CSDM estimates are presented.
