Classification of battlefield ground vehicles based on the acoustic emissions

Abstract

The acoustic emissions of ground vehicles contain a wealth of information that can be used for vehicle classification, especially on the battlefield and for scenarios where optical/radar-based sensor systems are inhibitive. In this chapter, we first briefly review the signatures buried in the acoustic emissions of ground vehicles, and then show the time-variations and uncertainties inherent in the acoustic features that are caused by the variations of environmental conditions as well as the variations of the distance between the vehicle and the sensor system. Considering the difficulties in establishing precise mathematical models to describe these variations and uncertainties, we focus on the fuzzy logic rule-based classifiers (FL-RBC), and compare their performance against the Bayesian classifier. The uniqueness of our approach lies in the following. First, to facilitate prompt decision making, the acoustic features were extracted from short time (about one second) intervals in which the acoustic measurements can be assumed to be stationary. Second, the choice for the number of rules in the FL-RBC was rationalized by the information inherent in the classification problem regarding the natural models of the vehicles and terrain conditions. And, third and finally, interval type-2 FL-RBCs were constructed to take advantage of the capabilities of interval type-2 fuzzy sets in modeling unknown time-variations and uncertainties. We also present the results of the experiments to evaluate the performance of all classifiers. © 2010 Springer-Verlag Berlin Heidelberg.