Information Sharing Between Localization, Tracking, and Identification Algorithms

Algorithms for ground vehicle classification using acoustic measurements often use features based on the relative amplitudes of harmonic components. The average power measured at each harmonic frequency fluctuates due to scattering of the acoustic waves during propagation and additive noise at the sensor. In this paper, we use propagation models for frequencydependent scattering of acoustic signals to study the performance and limitations of ground vehicle classifiers that use harmonic amplitudes as the features. The harmonic amplitudes for a given vehicle may vary due to other factors besides scattering, such as engine load and speed (revving) and the orientation of the source with respect to the sensor (aspect angle). Information about vehicle speed and aspect angle are available from localization and tracking algorithms in a networked sensor system. We evaluate the potential for improved classification accuracy when additional features are used from the localization and tracking algorithm. Examples from measured data are included to illustrate the results.