Optimal waveform selection for radar target classification

We apply a sequential experiment design procedure to the problem of signal selection for radar target classification. Radar waveforms are designed to discriminate between targets possessing a doubly-spread reflectivity function that are observed in clutter. The waveforms minimize decision time by maximizing the discrimination information in the echo signal. Each waveform selected maximizes the Kullback-Leibler (1951) information number that measures the dissimilarity between the observed target and alternative targets. We discuss two scenarios. In the first scenario, the target environment is assumed fixed during illumination. In this case the optimal waveform selection strategy leads to a fixed library of waveforms. During actual classification, the sequence in which the waveforms are selected from the library is determined from the noise to clutter power in the range-Doppler support of the targets. In the second scenario, the target environment changes between pulse transmissions. In this case, the maximum discrimination information is obtained by a repeated transmission of a single waveform designed from the reflectivity function of the targets. We show that our choice of signals can produce significant gains in detection performance.