Target coherence analysis using canonical correlation decomposition for SAS data

This paper analyzes the connection between the canonical correlations of sonar signals captured using two linear hydrophone arrays and the spatial coherence of the sources that are observed. Analysis of this connection allows for the use of a canonical correlation decomposition (CCD) framework for performing synthetic aperture sonar (SAS)-like processing. In particular, it can be used to extract correlation features of the targets in the frequency domain, which can then be used as a measure for target detection and classification. Results on real and simulated data indicate that coherence patterns are different in the presence of a target when compared to background clutter, and this pattern also changes as a function of the number of sensors on the array.