Defence Res. Establishment Atlantic, P.O. Box 1012, Dartmouth, NS, Canada
Noise normalization techniques for beamformed towed array data attempt to remove the ambient and self-noise directionality characteristics associated with the measurements in order to enhance detection of weak signals. The problem of normalization in bearing space for broadband or narrow-band plane wave detection is similar to normalization for signal detection in frequency domain. As a result, noise normalization techniques for the spatial domain may be modified versions of those used in the frequency domain [i.e., Struzinski, J. Acoust. Soc. Am. 76, 1738--1742 (1984)]. The aim of this paper is to develop a new concept of noise normalization for beamformed towed array data. This concept has two sets of operations, the first applies low angular resolution beamforming to obtain stable noise background estimates that retain the directionality characteristics of the noise field while the second includes the estimation of the geometric mean over a number of time snapshots in order to remove any high-level signal from the beam-power towed array estimates. In situations where the above two steps are insufficient to remove a strong broadband signal's influence from the noise background estimates, a third step may be applied for signal suppression. In principle, the above set of operations are similar to those used in frequency domain noise normalization techniques. Their numerical implementation, however, is fundamentally different. Performance analysis and comparison of the proposed normalization technique with the two pass split-window normalizer have been carried out by using synthetic and real data sets. Application results indicate similar performance for these two fundamentally different normalization schemes.