Ground Vehicle Classification by Acoustic Emission Exploitation

The makeup and location of hostile forces can be identified for battlefield surveillance or target acquisition by deploying a network of autonomous acoustic sensors. Each sensor would listen for acoustic emissions from ground vehicle traffic. Such emissions would then be analyzed by pattern recognition algorithms to determine the presence of acoustic signatures characteristic of potentially hostile forces. Conventional pattern recognition algorithms fail to provide reliable vehicle classifications at low signal-to-noise ratios or when there is interference from other vehicles. Information Systems Laboratories (ISL), proposes to investigate feature extraction algorithms that provide greater sensititity and greater immunity to interference so that reliable classifications can be produced under realistic operating conditions. The technology developed under this contract can be applied to monitoring highway traffic, such as determining the density of truck or bus traffic. Another application is predictive maintenance of rotating or reciprocating machinery. The algorithms can be adapted to distiguish signatures from "healthy" machinery from those of machines with early signs of failure.