Event identification by acoustic signature recognition Page: 4 of 9
This article is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to UNT Digital Library by the UNT Libraries Government Documents Department.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
Many security functions can be facilitated by hardware that classifies signatures at audio
frequencies. An obvious application is the covert monitoring of activity at remote airstrips in operations
such as counter narcotics.2 Signature classification hardware can also be used to detect breaches in secure
facilities; it can be programmed to detect events such as gunshots, while not being confused by similar
events such as thunder. In prisons, small, unobtrusive and inexpensive devices could be used to monitor
weak points in security for distinctive sounds such as digging or fence climbing. In access control, a gate
security system can be programmed to grant access only to a vehicle that emits an allowable acoustic
signature. Other transportation security systems can be based on audio-frequency signature
identification.
Pattern recognition hardware has not been widely used because several inherent problems have
remained unsolved. The most difficult technical problem is the identification of a suitable feature space.6
A feature space is a mathematical space in which the attributes of samples of a given class occupy a
limited region, while attributes of samples of other classes occupy other limited, yet distinct, regions.'
Although the Fourier frequency domain is often used in acoustic signal analysis, it is not a good
feature space for classifying acoustic signatures. The discrete Fourier transform washes out time
variations in the spectrum and introduces artifacts into the spectrum that are not present in the underlying
signal.' In contrast, the wavelet transform resolves a signal into both scale and time components and
provides good localization in both dimensions.' Wavelet scale is much easier to treat mathematically than
is Fourier frequency. Wavelet time-scale space is highly successful as a feature space for acoustic
signature classification.
The other problem inherent in pattern recognition is realization in inexpensive hardware. We have
built and operated several prototype hardware wavelet engines using off-the-shelf digital signal processing
(DSP) chips. The wavelet engine is nothing more than a set of finite impulse response (FIR) filters
arranged in a straightforward structure.'" Therefore, it is feasible to replace the DSP chips with dedicated
FIR chips," resulting in a wavelet engine that is small, inexpensive to produce in quantity, and simple
to program.
2. COMPLETED WORK
2.1 Airport Monitor
Typical results of wavelet analysis of airplane acoustic signatures are shown in Fig. 1. Time
series acoustic signatures for four different airplanes taking off from McGhee Tyson Airport (Knoxville,
Tennessee) are shown projected onto the eighth level of a 12-level Daubechies wavelet. The first level
Saab JetstreamFig. 1. Wavelet signatures of four airplanes in two classes.
Upcoming Pages
Here’s what’s next.
Search Inside
This article can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Article.
Dress, W. B. & Kercel, S. W. Event identification by acoustic signature recognition, article, July 1, 1995; Tennessee. (https://digital.library.unt.edu/ark:/67531/metadc623390/m1/4/: accessed April 10, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.