Machine learning-based channel estimation for insufficient redundancy OFDM receivers using comb-type pilot arrangement

Abstract

A severe drawback in broadband communications is the inter-symbol interference (ISI) originating from multipath fading, where one widely used solution is the orthogonal frequency-division multiplexing (OFDM) system. OFDM employs a block transmission, giving rise to inter-block interference (IBI) that can be remedied by using redundant elements. The standard solution is to insert a cyclic prefix (CP), whose length is equal to the channel order, and a set of pilots to estimate the channel, consuming, in part, the budgeting spectrum. This work proposes a machine learning (ML) based channel estimator for OFDM receivers operating with reduced redundancy and pilots. Our results confirm that the ML-designed receivers can achieve competitive bit-error-rate (BER) performance, opening new venues to improve spectrum utilization.