Complexity suppression of neural networks for PAPR reduction of OFDM signal and its FPGA implementation

In this paper, a neural network (NN) for peak power reduction of orthogonal frequency-division multiplexing (OFDM) signals is improved in order to suppress its computational complexity. Numerical experiments show that the proposed NN has less computational complexity than the conventional one. The number of IFFT in NN can be reduced to half, and the computational time can be suppressed by 32.7%. From the HDL simulation for FPGA implementation, hardware resouces are approximately suppressed by about 30%.