Deep Learning Based PAPR Reduction in 4G Mobile Communications

Deep Learning Networks are swiftly gaining notable popularity in the area of 4G communication domains and promise to offer emerging solutions for signal processing problems. An increase in Peak-to-Average Ratio (PAPR) is the critical problem of the 4G based OFDM communication system. As a result, improving the PAPR reduction potential of the OFDM systems with an optimal level of bit error rate of the system becomes a mandate. In this paper, a neural network model using CNN architecture is trained on the original OFDM symbols and built to decrease the PAPR value of the OFDM signals. Simulations conducted prove the outperformance of the designed system in terms of PAPR reductions with an acceptable level of BER degradations in OFDM signals in comparison to some of the conventional schemes.