面向高频段非理想信道的星座图智能优化设计方法

Intelligent optimization method for constellation design in high-frequency non-ideal channels

摘要:

随着通信系统不断向更高频段和更大带宽方向发展,硬件非理想性的影响愈发显著,尤其是功放非线性和相位噪声导致的信号畸变问题。为改善这一状况,提出了一种基于正交幅度调制(quadrature amplitude modulation, QAM)投影的人工智能辅助星座图设计方案。首先,借助基于离散傅里叶扩展正交频分复用(discrete Fourier transform spread orthogonal frequency division multiplexing, DFT-s-OFDM)驱动的神经网络模型,在非理想信道和时频性能指标损失等多重约束下进行星座学习,以降低系统误码率。随后,根据所提出的最小加权径角距离准则,将优化后的星座投影到QAM上,从而设计抗硬件非理想性强,且兼容现有体制的N-nMod_M-QAM星座。仿真结果表明,在相位噪声影响的信道下,提出的星座调制方案在性能增益和鲁棒性方面均优于同阶QAM、幅度相移键控(amplitude-phase shift keying, APSK)以及螺旋调制。

Abstract:

With the continuous development of communication systems towards higher frequency and larger bandwidths, the impact of hardware non-idealities, such as signal distortion caused by power amplifier nonlinearity and phase noise, has become increasingly significant. In an effort to address this challenge, this paper introduces an artificial intelligence-guided constellation design scheme based on quadrature amplitude modulation (QAM) projection. With the objective of reducing the bit error rate, a neural network model based on discrete Fourier transform spread orthogonal frequency division multiplexing (DFT s-OFDM) is utilized to facilitate constellation learning under multiple constraints, including non-ideal channels and time-frequency performance loss. Subsequently, the optimized constellation is projected onto QAM based on the proposed minimum radial-angular weighted distance, thus designing an N-nMod_M-QAM constellation that is resistant to hardware non ideality and compatible with existing systems. Simulation results show that under phase noise channel, the proposed constellation modulation scheme outperforms the same-order QAM, amplitude-phase-shift keying (APSK) and spiral modulation in terms of performance gain and robustness.