Improved Tone Reservation Method Based on Deep Learning for PAPR Reduction in OFDM System

Abstract

This paper utilizes deep learning (DL) in tone reservation (TR) to reduce the peak-to-average power ratio (PAPR) of orthogonal frequency division multiplexing (OFDM). We propose TR based on DL (DL-TR) algorithm by considering each iteration of the classical TR algorithm as a layer of a deep neural network (DNN), and then train the network offline to obtain the clipping ratio of each layer which can minimize the loss function that is the sum of PAPR and the increased transmit power. Compared with the conventional TR method, the simulation results show that the proposed DL-TR provides a better PAPR reduction and bit-error-rate (BER) performance.