A Low Complexity PAPR Reduction Scheme for OFDM Systems via Neural Networks

Abstract

Peak-to-average power ratio (PAPR) reduction is one of the key components in orthogonal frequency division multiplexing (OFDM) systems. Among various PAPR reduction techniques, artificial neural network (NN) has been one of the powerful techniques in reducing the PAPR due to its good generalization properties with flexible modeling and learning capabilities. In this letter, we propose a new method that uses NNs trained on the active constellation extension (ACE) signals to reduce the PAPR of OFDM signals. Unlike other NN based techniques, the proposed method employs a receiver NN unit, at the OFDM receiver side, achieving significant bit error rate (BER) improvement with low computational complexity.