Cycle-slip-tolerant decision-boundary creation with machine learning
Cycle-slip-tolerant decision-boundary creation with machine learning
复制标题
DOI：
10.1109/icp.2016.7510026
复制
发表时间：
2016-03
期刊：
影响因子：
--
通讯作者：
Hiroshi Kawase;Y. Mori;H. Hasegawa;Ken-ichi Sato
中科院分区：
文献类型：
--
作者：
Hiroshi Kawase;Y. Mori;H. Hasegawa;Ken-ichi Sato
使用支持向量机（SVM）等机器学习方法创建的最优符号决策边界可以在存在非线性信号失真的情况下提高系统性能。然而，如果接收到的边界生成训练信号包含由激光相位噪声引起的周期滑移，则该方法将失败。本文提出了一种新的容忍周期滑动的决策边界生成算法。我们提出的方案将旋转对称星座符号分组，并通过监测同一组符号之间的相位差来检测周期滑移。通过数值分析，我们证实了我们所提出的决策边界创建技术是免疫周期滑移的。
Optimum symbol-decision boundaries created using machine learning approaches such as support-vector machine (SVM) can enhance system performance in the presence of nonlinear signal distortion. However, they fail if the received training signals for boundary creation include cycle slips induced by laser phase noise. In this paper, we propose a novel decision-boundary generation algorithm that is tolerant to cycle slips. Our proposed scheme groups rotationally symmetric constellation symbols and detects cycle slips by monitoring the phase differences between symbols in the same group. Through numerical analysis, we confirm that our proposed decision-boundary creation technique is immune to cycle slips.