Radar emitter classification using intrapulse data

We consider the situation where a radar intercept receiver collects incoming pulse samples from a number of unknown emitters. Our objectives are to (1) determine the number of emitters present (cluster validation); (2) classify the incoming pulses according to the emitters from which they originate (clustering). The determination here is only based on intrinsic pulse shapes, without any inter-pulse information such as pulse repetition intervals, directions of arrival, carrier frequencies, or Doppler shifts. After pre-processing the received pulses and formulating the problem as multivariate clustering, we develop a coding length based on the Minimum Description Length (MDL) criterion for cluster validation and develop an efficient algorithm for clustering. Comparative study and computer simulations show that the performance of our method is very promising.