Performance Monitoring of Digital Communication Systems Using Extreme-Value Theory

Extreme-value theory has been suggested by various investigators as an efficient tool for use in monitoring the performance of digital communication systems. One drawback to such a procedure is that it requires learning periods to estimate certain parameters of the extremal distributions. Clearly, whenever an update of the estimates is needed, transmission of information must temporarily cease since a new learning period must begin. To avoid this problem, a technique is proposed for obtaining estimates of the parameters when the learning period is eliminated (i.e., estimates are obtained using noisy samples), and these parameter estimates are in turn used to estimate the probability of error of the communication system. The accuracy of this procedure is tested by computer simulation, and it is indicated how one can verify (in a statistical sense) whether or not this technique is appropriate for any given system.