Improving performance for matched field processing with a minimum variance beamformer

Abstract

Matched field processing (MFP) employing reduced minimum variance beamforming (RMVB) has been described in the literature as being robust to signal phase errors for vertical line arrays in modal noise. These techniques are extended to horizontal line arrays and evaluated for both horizontal and vertical arrays. The evaluation of RMVB is also more general in that it is extended to white and modal noise fields in the presence of phase or amplitude errors. Two phase error models were employed, one corresponding to constant sensor-position errors the other to short-term fluctuations of mode phase during the covariance matrix estimation. For minimum variance beamforming (MVB), in the presence of both types of random phase errors, array gain, peak-to-sidelobe ratios, and peak-to-background ratios all decreased in a modal noise field. Performance was far less sensitive to phase errors in spatially white noise than in modal noise. MVB was modified by reducing the number of eigenvectors employed in the matching. For the modified forms performance was improved over that of MVB with an attendant reduction in the number of computations. The equispaced horizontal arrays consistently outperformed and were more robust than the equispaced vertical arrays. Similar results were obtained with amplitude errors that were constant during the estimation.