A Novel Tone Reservation Scheme Based on Deep Learning for PAPR Reduction in OFDM Systems
Benwei Wang, Qintuya Si, Minglu Jin
Abstract
A major defect of orthogonal frequency division multiplexing (OFDM) systems is the high peak-to-average power ratio (PAPR). In this letter, a novel scheme based on deep leaning, called tone reservation network (TRNet), is proposed for OFDM systems to improve the performance of the tone reservation (TR) technique. More specifically, TRNet reserves a part of tones to generate a peak-canceling signal. The feedforward neural network is used to adaptively generate a peak-canceling signal according to the characteristics of the input signal. Computer simulation results show that the proposed scheme provides a better PAPR reduction performance with fewer reserved tones, which is also beneficial to improve the bandwidth efficiency.