On M-ary Sequential Hypothesis Testing for the Classification of Radar Signals

This paper is concerned with the performance of M-ary sequential hypothesis tests applied to the classification of radar signals. This investigation is primarily motivated by interest in the design of radar target identification (RTI) systems, where the decision algorithm produces the identity, among M possibilities, based upon several observations of vectors of signal parameters. The performance of several M-ary sequential algorithms are considered. These include the parametric algorithms due to Reed, Palmer, and Armitage, as well as a sequential version of the nonparametric nearest neighbor algorithm. The average number of required measurements and the resulting error performance, of these techniques are compared, and modifications of the algorithmis are considered.