Target Identification Using Multifrequency Radar Sensor Networks

We present techniques for target identification using resonance-region, multifrequency radar sensor networks. The majority-vote (MV) and sum-distance (SD) nearest-neighbor (NN) algorithms are used. The NN reference set initially contains samples of target features over the possible ranges of target aspect angles. We use a data condensation rule to condense the initial reference set. Simulation results show that the identification error probabilities can be significantly lowered by 1. increasing the number of radar sensors, 2. increasing the number of frequencies, 3. using the complex features instead of the amplitude ones, 4. using the SD algorithm instead of the MV one.