A Novel PAPR Reduction Scheme Based on Deep Autoencoder Network for FBMC Systems

Filter bank multicarrier (FBMC) is a crucial complementary waveform to orthogonal frequency-division multiplexing (OFDM) in future communication systems. However, FBMC systems also suffer from the drawback of an excessively high peak-to-average power ratio (PAPR). Moreover, PAPR reduction methods designed for OFDM systems cannot be directly applied to FBMC systems due to the overlapping nature of FBMC symbols. In this paper, we propose a novel deep learning (DL)-based PAPR reduction scheme for FBMC systems. This innovative approach employs a deep denoising autoencoder (DAE) network in the time domain to suppress the PAPR of FBMC signals and reconstruct the ideal signal. Simulation results demonstrate that the proposed DAE-PAPR scheme achieves a PAPR reduction gain of approximately 2.3 dB to 4.5 dB compared to conventional state-of-the-art methods, while maintaining excellent bit error rate (BER) performance and out-of-band energy leakage characteristics.