A Neural Network Based Method for Peak-to-Average Power Ratio Reduction in Orthogonal Frequency Division Multiplexing Systems

Orthogonal Frequency Division Multiplexing (OFDM) technology, renowned for its multicarrier transmission paradigm, encounters a formidable challenge characterized by an elevated Peak-to-Average Power Ratio (PAPR). This phenomenon can precipitate an escalation in signal distortion. In this study, we present an innovative approach wherein a neural network is harnessed to enhance the attenuation of PAPR, synthesizing the techniques of peak clipping and dilated convolution. Simultaneously, a recovery module is integrated to optimize the signal quality post-suppression, especially at the receiver's terminal. Our empirical findings manifest a noteworthy upswing in the efficacy of PAPR suppression, underscoring the remarkable performance augmentation realized through the proposed methodology.