An Optimized Deep Neural Network with Heuristic Algorithm for PAPR Reduction Model in OFDM Communication Systems

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) is a flexible method utilized in several wireless and wired standards because of its poor complexity receiver model and robustness in multipath propagation environments. The OFDM communication model attains superior data rates and spectral efficiency for mobile wireless communication systems. Although this scheme provides superior Quality of Service (QoS), it attains a high Peak-to-Average Power Ratio (PAPR), which is considered as a complicated problem and also needs to be determined in the OFDM scheme. The PAPR creates distortion in the transmitted signal because of the power amplifier’s nonlinearity. As a result, the PAPR minimization of the OFDM model is one of the significant experimental topics in OFDM and has been the topic of detailed examination by numerous experts. Though there have been various PAPR reduction techniques in the past years, the models can’t decrease the system’s Bit Error Rate (BER) because of the signal distortion. Therefore, a new strategy is required for the PAPR reduction in the OFDM system. In this paper, a new PAPR reduction strategy is implemented with the help of Optimized Deep Neural Networks (ODNN), where the parameters of DNN like hidden neuron count and epochs are optimized through a new Improved Billiards-inspired Optimization Algorithm (I-BOA). This suggested framework is employed to determine the constellation demapping and mapping in each subcarrier. The major goal of the developed framework is to reduce the PAPR and BER in the communication system. Different validations are executed in the developed framework to display the efficiency of the PAPR and BER. While considering PAPR, the findings of the developed I-BOA-ODNN method achieve 17.7%, 19.4%, 16.5%, and 14.6% than Genetic Algorithm (GA), Particle Swarm Algorithm (PSO), Artificial Bee Colony Algorithm (ABC), and Billiards-inspired Optimization Algorithm (BOA) in terms of the mean. Throughout the validation, the offered I-BOA-ODNN model performs superior to the existing heuristic algorithms.