Automated Vehicle Identification at a Distance

Vehicle identification can be effected by acquiring multi-spectral signatures (i.e., acoustic and electromagnetic emissions) in an analogous fashion to the conventional biometric triad of fingerprints, facial recognition, and speech analysis used for identification of people. The acoustic metrics would include emissions unique to the engine and exhaust system for a particular vehicle, while the electromagnetic metrics would include RF harmonics radiated from the vehicle when illuminated by an RF transmitter. The spectrographic content of these signatures will be used to define metrics such as a multi-dimension vector composed of cepstral coefficients and their derivatives. Specifically, the offerer will develop a sensing platform that will incorporate microphones and harmonic RF transceivers to acquire long range acoustic and RF harmonic data from representative vehicles. The data will be analyzed to develop frequency domain metrics for vehicle identification. Methods will also be developed to determine search and classification algorithms that will permit vehicle identity verification based upon a database of vehicle RF harmonic and acoustic signatures created for this purpose. An equally important research goal addresses the development of a component-based framework to provide the acquisition and analysis capability for the spectral emissions. This reconfigurable design will provide future researchers in the field with an infrastructure that will allow easy interface with third party hardware and analysis algorithms. Proof of concept will be demonstrated by showing how these metrics can be used to differentiate between vehicles.