A joint optical signal-to-noise ratio (OSNR) monitoring and modulation format identification (MFI) is proposed based on signal amplitude histograms (AHs) by using convolutional neural network (CNN). By the use of the proposed scheme, the OSNR monitoring accuracy can be achieved 97.6%, 100%, and 100% for 28GBaud QPSK, 8-QAM, and 16-QAM respectively within a wide OSNR range (10–26 dB). Similarly, the MFI accuracy can all achieve 100% for the three modulation formats. In addition, the effect of the network structure on the CNN performance is also studied. In the proposed scheme, OSNRs and modulation formats can be simultaneously identified in high accuracy by using a single CNN without additional hardware assistance, which should be attractive for cost-effective multiple parameters estimation in future dynamic optical networks.

利用卷积神经网络（CNN），基于信号幅度直方图（AHs），提出了一种联合的光学信噪比（OSNR）监测和调制格式识别（MFI）。通过使用所提出的方案，在宽的OSNR范围（10–26 dB）内，分别对28GBaud QPSK，8-QAM和16-QAM可以实现97.6％，100％和100％的OSNR监视精度。同样，三种调制格式的MFI精度都可以达到100％。此外，还研究了网络结构对CNN性能的影响。在提出的方案中，可以通过使用单个CNN以高精度同时识别OSNR和调制格式，而无需额外的硬件帮助，这对于将来的动态光网络中具有成本效益的多参数估计应具有吸引力。