Classifying acoustic signals detected by distributed sensor networks is a difficult problem due to the wide variations that can occur in the transmission of terrestrial, subterranean, seismic and aerial events. An acoustic event classifier was developed that uses particle swarm optimization to perform a flexible time correlation of a sensed acoustic signature to reference data. In order to mitigate the effects from interference such as multipath, the classifier fuses signatures from multiple sensors to form a composite sensed acoustic signature and then automatically matches the composite signature with reference data. The approach can classify all types of acoustic events but is particularly well suited to explosive events such as gun shots, mortar blasts and improvised explosive devices that produce an acoustic signature having a shock wave component that is aperiodic and non-linear. The classifier was applied to field data and yielded excellent results in terms of reconstructing degraded acoustic signatures from multiple sensors and in classifying disparate acoustic events.

Date Published: 22 May 2013
PDF: 11 pages
Proc. SPIE 8742, Ground/Air Multisensor Interoperability, Integration, and Networking for Persistent ISR IV, 87420V (22 May 2013); doi: 10.1117/12.2018250

Published in SPIE Proceedings Vol. 8742:
Ground/Air Multisensor Interoperability, Integration, and Networking for Persistent ISR IV
Tien Pham; Michael A. Kolodny; Kevin L. Priddy, Editor(s)