A Neural Network-Enabled OTFS-PAPR Reduction with Low Computational Complexity

Abstract

This study proposes a new solution to overcome the high peak-to-average power ratio (PAPR) in Orthogonal Time Frequency Space (OTFS) by using an Artificial Neural Network (ANN) algorithm. The algorithm checks the magnitude (power) of each element in the matrix of the first stage of the inverse symplectic finite Fourier transform (ISFFT) process against a pre-specified threshold and, consequently adjusts the elements whose magnitudes exceed the threshold. This is achieved by using the ANN algorithm to apply fractional shifts to the elements of the original delay-Doppler (DD) data matrix without changing their orientation. The simulation results demonstrated a significant PAPR reduction while maintaining the system performance in terms of the Bit Error Rate (BER), with almost the same computational complexity of the conventional OTFS system.