Classification of Military Ground Vehicles Using Time Domain Harmonics' Amplitudes

In the context of the United Nations peacekeeping operations, we developed an energy-efficient method for the detection and classification of military vehicles using a group of wireless sensors. The method is adequate for low-power unattended sensors which perform sensing, feature extraction, and classification in a standalone scenario. Harmonics' amplitudes approximating the harmonic signature of the time domain acoustic signal captured by wireless sensor nodes are estimated for vehicle discrimination. The computational complexity for the time domain features extracted from ground vehicles' acoustic signals is lower than their equivalent spectral features. Classification is performed using a multilayer feedforward neural network, where discrimination between vehicles depends on their acoustic signature irrespective of their speed or location. Evaluation of the time domain method, through processing of an acoustic data set for heavyweight and lightweight military ground vehicles with comparison to spectral features, shows that time domain harmonics' amplitudes are simpler to obtain and provide the reliability of the spectral features in both the detection and false alarm rate.