On the Performance of Some Sequential Multiple-Decision Procedures

The performance of several sequential procedures for the following multiple-decision problem is investigated. Samples from k random processes (or populations) are available, k at a time (one from each process), to a receiver or data processor. One process contains a signal; the other k - 1 are statistically identical noise. The receiver is to select the odd process (locate the signal), with prescribed probability of error. The optimal receiver makes the selection in minimum average time. Analytical and Monte Carlo computations were performed under the hypothesis that the processes sampled are Rayleigh; however, a method for extrapolating results to other cases is given. The parameter k is allowed to vary from 2 to 1000.