| 摘要: |
| 复杂通信场景下非高斯干扰噪声使得传统多输入多输出（Multiple-Input Multiple-Output，MIMO）检测的误码性能受限。为解决在相关动态干扰噪声环境下的MIMO信号检测问题，提出了一种将半正定松弛（Semi-definite Relaxation，SDR）检测器和条件生成对抗网络（conditional Generative Adversarial Network，cGAN）联合的新型迭代检测算法，称为SDR-cGAN。SDR对接收信号进行初始检测， cGAN实现对相关干扰噪声的精确估计，通过两者构成的迭代框架，不断消除干扰噪声残差，从而提升检测算法的误码性能。仿真实验结果表明，当系统收发天线数相等，信噪比为15 dB时，SDR-cGAN算法误码率接近10-6，与其他基于深度学习的检测算法相比，误码率降低1~2个数量级，表现出接近最优检测的优良性能。在不同的调制方式下，SDR-cGAN算法较传统检测算法的误码性能也展现出明显优势，具有较好的鲁棒性。 |
| 关键词: 多输入多输出 信号检测 相关干扰 半正定松弛 条件生成对抗网络 |
| DOI：10.20079/j.issn.1001-893x.240513003 |
|
| 基金项目: |
|
| MIMO Signal Detection Algorithm Based on cGAN in Correlated Interference Environment |
| KANG Xiaofei,LI Yumei,LIANG Qiyue |
| (School of Communication and Information Engineering，Xi揳n University of Science and Technology，Xi揳n 710600，China) |
| Abstract: |
| Non-Gaussian noise interference in complex communication scenarios severely limits the error performance of traditional multiple-input multiple-output(MIMO) detection.In order to solve the problem of MIMO signal detection in the context of relevant dynamic noise interference,a novel iterative detection algorithm combining the semi-definite relaxation(SDR) detector and conditional generative adversarial network(cGAN) is proposed.It搒 called SDR-cGAN.The SDR performs initial detection of the received signal,and the cGAN network realizes accurate estimation of relevant noise interference.Through the iterative framework composed of SDR and cGAN,noise interference residuals are continuously eliminated,thus improving the error performance of the detection algorithm.The simulation results show that when the number of antennas is equal and the signal-to-noise ratio(SNR) is 15 dB,the bit error rate of SDR-cGAN algorithm is close to 10-6.Compared with that of other detection algorithms based on deep learning,the bit error rate of SDR-cGAN algorithm is reduced by one to two orders of magnitude,showing excellent performance close to optimal detection.Under different modulation modes,SDR-cGAN algorithm also shows obvious advantages in error performance compared with traditional detection algorithms,and has better robustness. |
| Key words: multiple-input multiple-output signal detection correlation interference semi-definite relaxation conditional generative adversarial network |
《电讯技术》编辑部 