Optimization of time-domain header processing filters using genetic algorithms

We present the use of a genetic algorithm as an optimization method for the design of header recognition filters based on time-delay lines. The discrimination of the target pattern against all other possible bit patterns represents the fitness parameter used in the selection process. For phase-shift-keyed header patterns, filters generated by the algorithm exhibit improved discrimination compared to matched filters. Time-domain filters providing non-zero discrimination while exhibiting no ambiguities are also found for specific amplitude-shift- keyed headers. Delay-line-based time-domain filters can be implemented in a planar lightwave circuit platform by simple phase and amplitude adjustments on the time delay taps.