Optical Signal-to-Noise Ratio (OSNR) is an important criterion for evaluating the performance of communication systems. For optical fiber communication systems, stable operation can only be ensured if the OSNR at the receiver end exceeds the receiver's OSNR receiving tolerance. The larger the value, the better the system stability. Therefore, monitoring the OSNR value of the transmission system is very important, and real-time and accurate monitoring of OSNR has always been a challenge in optical fiber communication systems.

To achieve a simple, efficient, and high-precision OSNR monitoring, based on the mean and variance of the signal amplitude histogram, combined with a Deep Neural Network (DNN), the article proposes a highly accurate OSNR monitoring method. This method utilizes the mean and variance of different regions of the received signal amplitude histogram to extract correlation features of the signal amplitude histogram as it changes with OSNR using the DNN, thereby enabling the monitoring of the system's OSNR.

This monitoring scheme is used to experimentally monitor the receiver OSNR of Quadrature Phase Shift Keying (QPSK) /16 Quadrature Amplitude Modulation (QAM) /32QAM transmission systems with three different modulation formats. The results show that the maximum standard errors of the estimated and actual OSNR values of the three different modulation formats in the back-to-back transmission system are 0.048, 0.140, and 0.180 dB without fiber transmission. After passing through 2 000, 800 and 800 km optical fiber transmission, the maximum standard errors of the estimated and actual OSNR values of the receiving end of the three different modulation formats transmission systems are 0.082, 0.250 and 0.630 dB, respectively. Compared with the non fiber transmission system, the maximum error of the estimated OSNR values of the system after optical fiber transmission does not exceed 0.5 dB. This indicates that the OSNR monitoring scheme can achieve precise monitoring of transmission system performance and is not affected by fiber optic lines.

The DNN OSNR monitoring method based on the mean and variance of the signal amplitude histogram proposed in this paper has been experimentally verified. The maximum error between this scheme and the actual system receiver OSNR does not exceed 1dB, indicating that this scheme can accurately predict the system receiver OSNR value.