Deep Learning-Based PAPR Suppression of OFDM Signals with Clipping Constraint

This study proposes a method to improve suppression performance by introducing a clipping-constraint into the loss function of PRNet, a deep learning model suitable for PAPR suppression in OFDM signals. Numerical experiments have confirmed that while maintaining BER performance equivalent to that of normal OFDM signals, this method improves PAPR performance by approximately 1.5 dB compared to the conventional method and about 4.5 dB compared to normal OFDM signals.