Genetic Algorithm Based Optimization of Encoding Sequence for a Reduced Complexity OFDM Time Synchronization Technique

A new Genetic Algorithm (GA) based sequence optimization is here proposed. The studied sequence, which we refer to as Encoding Sequence (ES), is used in a recently proposed OFDM preamble-based synchronization technique. The optimal ES is generated offline by differentially correlating the preamble samples to each other respecting a shift. To determine the frame start at the receiver, the ES is correlated to a version of the received signal generated using differential correlation respecting the same shift as in the ES generation. This treatment, carried online, is of high complexity due to the differential correlation operations. To reduce the computational load at the receiver, the idea here is to replace the differential correlation operations, carried during the synchronization process, by simple sign changes. To this aim, we need to derive the best sub-optimal ES, in terms of providing near optimal performance, from a finite QPSK alphabet (±1±j). Even if this search procedure is carried off-line, considering an exhaustive search over the whole ES candidates within QPSK alphabet results in a prohibitive complexity. As an alternative, we here suggest to use a GA-based approach. Performance comparison of the proposed ES, generated through GA-based search, to the optimal ES, generated from the preamble, and a previously proposed QPSK ES optimized through an iterative algorithm has been conducted. It is demonstrated that compared to the iteratively generated ES, the herein proposed one improves the detection accuracy. The GA-based provides slightly lower detection accuracy with respect to the optimal ES, yet with important reduction in the receiver complexity.