A Memetic Algorithm for Deinterleaving Pulse Trains

This paper deals with the problem of deinterleaving a sequence of signals received from different emitters at different time steps. It is assumed that this pulse sequence can be modeled by a collection of processes over disjoint finite sub-alphabets, which have been randomly interleaved by a switch process. A known method to solve this problem is to maximize the likelihood of the model which involves a partitioning problem of the whole alphabet. This work presents a new memetic algorithm using a dedicated likelihood-based crossover to efficiently explore the space of possible partitions. The algorithm is first evaluated on synthetic data generated with Markov processes, then its performance is assessed on electronic warfare datasets.

Automated summary

This paper addresses the problem of separating a sequence of signals received from different emitters at different time steps. It proposes a new memetic algorithm that uses a likelihood-based crossover to explore the space of possible partitions efficiently. The algorithm is evaluated on synthetic data generated with Markov processes and on electronic warfare datasets.