Share Email Print

Proceedings Paper

New classification results using temporal and spatial fusion
Author(s): Gervasio Prado
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

The promise of acoustic classification of vehicles is based on the expectations provided by the human ability to differentiate sounds from familiar vehicles. Some of these promises have not been fully achieved in practice, necessitating a "reality check" on the uses and limitations of the technology. Some of those limitations are: Operation in the real-world environment of outdoor propagation over rough terrain in the presence of natural and cultural background noise sources. A great number of the new applications are used against civilian-type vehicles that have emissions that are substantially lower than those of military vehicles. The success of speech recognition systems has also fueled some of these expectations. But before we take the analogy too far, we must note that there is a vast difference between the two tasks. Speech occupies a wider bandwidth than vehicle noise and is much more richly modulated than vehicle noises. Consequently there is much more information content to extract and more features to rely on than in the vehicle classification problem. Starting with a vehicle classification algorithm based on a neural network trained to recognize two different vehicles, we illustrate how by creating a continuous track on the target and integrating the output of the classifier over the life of the track we can improve the confidence of the classification results. Similarly fusing the results obtained by two or more sensors spread around the target can further improve the classification performance, cutting the rate of erroneous classifications by a third or more.

Paper Details

Date Published: 2 May 2006
PDF: 9 pages
Proc. SPIE 6231, Unattended Ground, Sea, and Air Sensor Technologies and Applications VIII, 62310N (2 May 2006); doi: 10.1117/12.673725
Show Author Affiliations
Gervasio Prado, SenTech, Inc. (United States)

Published in SPIE Proceedings Vol. 6231:
Unattended Ground, Sea, and Air Sensor Technologies and Applications VIII
Edward M. Carapezza, Editor(s)

© SPIE. Terms of Use
Back to Top