A Newral Network for reducing the Peak-to-Average Power Ratio in Orthogonal Frequency-Division Multiplexing Systems

点击翻译按钮获取中文摘要

Orthogonal frequency-division multiplexing(OFDM) modulation can reduce the influence of inter-symbol interference and enable high-quality communication. However, an OFDM signal has a large instantaneous peak power, which is measured as peak-to-average power ratio(PAPR), since the subcarrier signals are modulated independently. A variety of techniques for reducing PAPR have been proposed, the selective mapping method(SLM) is the simplest scrambing for reducing the PAPR in which it generates several scrambimg sequences at random and selects the sequence that gives the lowest PAPR. Although SLM has few computation time, the PAPR is not enough reduced.We have proposed the PAPR reduction method, in which the PAPR reducing problem is formulated as a combinatorial optimization problem and Hopfield neural nertwork(HNN) is applied to solving the optimization. HoweverHNN does not sufficiently improve the performance of conventional methods because the mechanism of HNN is the same as that of the gradient descent method, and if the state is caught in a local minimum point then HNN cannot escape from the point and does generate any novel solutions.In this research, we propose a novel PAPR reduction method by using the chaotic neural networks(CNN), which has been proposed by Nozawa and it has better performance for combinatorial optimization. First, we formulate the PAPR reduction problem as a combinatorial optimization problem, and HNN is introduced for the optimization. To improve the performance, a chaotic neural network is applied for leading to considerable gains, and we evaluate its performance by numerical experiments.