Signal Processing Applications for Submarines, Surveillance and Survival
Monday December 11, 2000, Afternoon
Adaptive signal processing and neural network based methodologies are topics of active
research within Australia’s Defence Science and Technology Organisation.
Their application to sonar systems and land-based acoustic surveillance systems is
considered here. For example, submarines have both hull-mounted and towed arrays
of hydrophonic sensors. The digital data from these sensors are processed to extract
information on acoustic signals propagating in the underwater environment.
Basic signal processing for passive sonar arrays involves applying the fast Fourier
transform in both the temporal and spatial domains. More advanced sonar processors
implement adaptive beamforming algorithms to maximise the output signal-to-noise ratio
of the sonar array in spatially correlated noise fields. Even at frequencies well below
the design frequency of the array, these optimal spatial filters are shown to enhance
spatial resolution, to minimise sidelobes, and to steer nulls in the direction of interfering
sources. However, adaptive beamforming can lead to signal suppression in the
presence of system errors caused by the phase or amplitude responses of the sensors
being mismatched, or by the knowledge of the sensor positions being imperfect.
During an ownship manoeuvre, this signal suppression property can be exploited to
refine the position estimates of the acoustic sensors in a long flexible array towed
behind the submarine. Passive sonar signal processing algorithms developed for
submarine applications can be modified to process microphonic sensor data for
acoustic surveillance of the land environment.
Unattended ground-based acoustic
sensing systems are often deployed in remote locations to detect, classify, localise and
track sources of military interest such as aircraft, ground vehicles and weapon fire.
Finally, the detection and classification of sea mines currently requires a purpose-built
naval vessel to enter the minefield so that the capability and operational performance
of the minehunter’s active sonar system are critical to mission success and survival.
Various signal and image processing algorithms, some of which feature neural networks,
are being developed for mine-hunting applications. These algorithms are described and
the results presented.