Sonar Processing Methods for Reverberation-Limited Undersea Environments


  • Ilkka Karasalo
  • Magnus Lundberg Nordenvaad
  • Bernt Nilsson
  • Per-Axel Karlsson
  • Elias Parastates

Publish date: 2009-01-08

Report number: FOI-R--2661--SE

Pages: 44

Written in: English


  • reverberation suppression
  • reverberation whitening
  • AR modelling
  • active sonar
  • synthetic sperture sonar
  • SAS
  • space-time adaptive processing
  • STAP


This report presents preliminary results from a study of promising methods with the potential to improve the performance of active sonar systems in reverberation-limited environments. First, two methods for whitening of reverberation noise have been implemented and studied using simulated and sea trial data. In the first method the matched filter receiver is tuned for each rangebin using an estimate of the reverberation covariance matrix. It is shown to provide better detection performance on a set of simulated data, but at a very large cost in the form of calculations. The second method uses autoregressive (AR) modelling of the reverberation noise, and the AR coefficients are then used to whiten the reverberation in sea trial data. Detection performance was improved significantly using the AR filter, and this was achieved at a fairly low computational cost. Second, Space-Time Adaptive Processing (STAP) is a promising signal processing technique that offers good mitigation of spatio-temporal interference, such as reverberation and jammers. Hence, it facilitates the detection of weak and slowly moving targets whose properties lie close to the interference characteristics. The major drawback of STAP, compared to conventional processing, is a substantial increase in degrees of freedom. This in turn yields large computational costs and a need of large training data sets which are not always readily available. This report proposes a hierarchical approach to overcome these shortcomings where in each step a decreasing number of sub-problems is solved. In this way, the complexity is greatly reduced compared to standard STAP approaches. Also, since the method combines solutions to sub-problems of smaller dimensionality, the required size of the noise training data set is also greatly reduced. As a result, the derived scheme performs better than standard STAP algorithms for small sample support. Third, Synthetic Aperture Sonar (SAS) processing, commonly used in advanced mine hunting sonars, is here applied in lower frequency surveillance applications. The SAS processing is added as a complement to the conventional single ping processing of the echoes received by a moving surveillance sonar, with the purpose to significantly increase the azimuthal (along-track) resolution in the sonar image. This could enable a more accurate and informative mapping of the surveyed scene, including higher resolution of multiple targets, detection of target shadows in the bottom reverberation, and resolution of the shapes of such shadows. Simulations made under idealized surveillance conditions, in particular constant (both space- and time-independent) sound speed and negligible errors in positioning of the sonar, indicate that good results can be obtained for relevant distances and frequencies. Less than ideal conditions are also investigated, including effects of errors in navigation data of the sonar and multiple bottom and surface reflections.