Page 59 - Fall_DTF
P. 59

Physics-Based Signal
Processing Approaches for
Underwater Acoustic Sensing
Lisa Zurk Physics-based approaches can greatly improve the classification and
Addmr localization of underwater sources in complex propagation environments.
Applied Physics Laboratory Intrnductinn
University ofwashington _ _ _ _ _ _ _
1013 NE 400‘ Sun‘ The undersea world is an important and fascinating domain, but it is also a very
S . challenging environment for sensing. The dominant modality for underwater
cattle, Washington 98105 _ _ _ _ _ _
USA sensing is acoustics because electromagnetic and optical waves attenuate quickly
in seawater. Fortunately, many sources of interest to humans, such as biologics
Email-' (whales, dolphins, fishes) and man-made vehicles (submarines, ships), have dis-
Z|l1'k1@|lW-Ed“ tinct acoustic signatures that can be detected passively or actively. Sound radiated
in the lower part of the acoustic band (i.e., 100s of hertz or below) can travel
over fairly long distances, and these signals can be detected at significant ranges.
However, the propagation over these distances can change the acoustic signature
appreciably due to the physics of the underwater propagation, which introduces
phenomena such as refraction and reflected multipaths.
Several of the most promising techniques in underwater signal processing used
in oceanography are based on exploiting knowledge of the expected structure of
the acoustic signature (in both the frequency and time domains). This structure is
a result of the radiated source pattern (for passive detection) or the target reflec-
tion pattern (for active detection) and the modulation imposed by propagation
through the acoustic channel. If the structure is known, a tirne-frequency filter
can be designed to extract sources of interest. Furthermore, with an appropriate
array aperture, a filter or bearnformer can be constructed based on the expected
spatial structure.
Here examples of physics-based spectral, temporal, and spatial filters are present-
ed and demonstrated for biological and man-made sources in both shallow and
deepwater environments.
Time and Frequency Structure
Acoustic signals can be represented in both the time and frequency domains. hi
the time domain, changes in acoustic amplitude correspond to the time-varying
strength of the acoustic signal as measured at the receiver. As an example, con-
sider Figure 1, left, that shows the sound amplitude measured with a microphone
of a voice recording from the song “That’s Amore” over a five-second period. The
amplitude clearly varies as the singer articulates, but the tonal components or
the frequency content is also varying (i.e., the notes in the song). To extract the
frequency signature, a short-term Fourier tra.nsform can be applied with a sliding
window. The window needs to be small enough so that the assumption of station-
arity is valid, that is, the frequency does not vary appreciably in that short window
(Van Trees and Bell, 1968). This time-frequency output, defined as a spectrogram,
is seen in Figure 1, right. It can be seen that the frequency content is time varying,
pertaining to the changing harmonic structure of the song.
©2018Acousz1coISOci'ezy ofAmm'm. All rights reserved. voIume14,i'ssue3 1 Fall 2013 | Acoustic: Today | 57

   57   58   59   60   61