Acoustic Signature Detection Idea
The idea of using acoustic signatures to discriminate anti-personnel
landmines from ground clutter was discussed (along with other methods)
in a 1996 Jason study (sponsored by DARPA and led by Paul Horowitz).
The following passage is an excerpt from that report:
3.3 Acoustic Signature
A typical antipersonnel mine is comparable in size and shape to a
hockey puck. It is cylindrical, with a plastic casing typically
1/8" thick, containing a cast explosive charge
that fills the majority of its internal space. There is a small
explosive primer, fired mechanically by a pin and spring
arrangement, which includes some air spaces. There may be a
small gap between the explosive and casing, caused by shrinkage.
There is usually a substantial air space above the explosive
charge, surrounding the trigger and detonator mechanism.
This object has different acoustic properties than the
surrounding soil. It is of known size, composition, and weight.
It is a highly symmetric object, consisting mostly of homogeneous
materials. As an object in isolation it is guaranteed to have a
characteristic set of resonances -- bumps and dips in its
response to an applied acoustic sweep through a range of
frequencies whose wavelengths are comparable to the size of the
object. We can think of this as an acoustic "signature." For
a homogeneous object (like a hockey puck) one can, of course,
calculate the acoustic modes from first principles. For an
object with the internal complexity of a mine one would simply
measure a sample, in an acoustic setup consisting of a driven
transducer and an acoustic detecting transducer. The modal
response will, in general, depend on the location of driver and
detector; but the overall pattern will show invariant spectral
features.
The idea, of course, is to measure the acoustic response of the
candidate object, in situ, to determine if it is a mine.
This is best done with a contacting probe -- we can think of it
as an enhanced demining prod -- which might be built as a closely
spaced pair of elastomeric tips, one driving and the other
detecting. The range of acoustic frequencies that should be used
is roughly 5 kHz - 50 kHz (corresponding to wavelengths in plastic
of 40 cm - 4 cm).
The success of such a method depends upon being able to see the
acoustic signature of a mine, when it is embedded in the
heterogeneous matrix of soil, rocks, etc. The simplest
procedure is simply to try the measurement -- place some mines in
typical soil, then measure the swept acoustic response (or,
equivalently, the Fourier Transform of its response to a short
impulse).
We do not know of any such data. However, from rather simple
arguments we believe that resonant frequencies of order a few
kilohertz, with resonant Q's (quality factor) of order 10--100,
will characterize the airspace resonances, and that these should
be easily detectable with contacting probe pairs. For technical
reasons (having to do with separation of stimulus and response
signals) it is probably better to use an impulse rather than a
swept frequency source. With a pulsed source the detector sees
only the decaying response, and does not have to contend with the
stronger source signal itself. In practice one would generate a
train of pulses, perhaps 10-100 per second, and use standard
signal averaging techniques to improve the sensitivity and
accuracy.
The solid-body resonant modes are probably highly damped by
contact with the soil, and probably not useful. However, the
internal structures of the mine -- the explosive compartment, the
washers, actuators, detonator, booster, etc. -- will also
exhibit resonances unless they are firmly bonded to the outer
structure. An additional possibility is resonances in small
elastic structures such as springs and struts.
If acoustic signature analysis is practical, its implementation
should be extremely simple -- an acoustic transducer pair
(perhaps a PVDF polymer), with a low-power swept oscillator or
pulse generator, and audio amplifier/detector, followed by some
backend processing to compare the measured response with a
library of premeasured responses. We estimate such technology
should cost less than $1000 each in moderate quantities, and
weigh no more than a few pounds, including signal processing and
batteries (we were told that a medical ultrasound transmitter and
Doppler detector, including processing electronics, costs $800).