研究目的
To investigate the classification methods of SVM machine learning detection with performance comparisons for QAM modulated optical interconnection.
研究成果
The experimental results indicate that the SVM methods can well detect the nonlinear damage of signal and classify the signal reasonably. The SVM multi-classification based on the in-phase and quadrature component has the lowest complexity while maintaining a certain BER performance.
研究不足
The study is limited to QAM-DMT optical transmission link based on the Mach-Zehnder modulator and 10-km standard single mode fiber. The performance of SVM methods may vary with different modulation formats and nonlinear distortions.
1：Experimental Design and Method Selection:
Implemented SVM detection in a QAM-DMT optical transmission link based on the Mach-Zehnder modulator (MZM) and 10-km standard single mode fiber (SSMF). Four multi-classification detection methods based on SVM were investigated.
2：Sample Selection and Data Sources:
QAM-DMT signal with aggregated data rate of 100-Gbps (with 128 carriers) was generated.
3：List of Experimental Equipment and Materials:
MZM: Mach-Zehnder Modulator; PD: Photodiode; AWG: arbitrary waveform generator; EA: electrical amplifier; DSO: digital storage oscilloscope; SSMF: standard single mode fiber; PC: polarization controller.
4：Experimental Procedures and Operational Workflow:
The received optical power was set at 4 dBm. The example constellation diagram was the 17th carrier for optical B2B case.
5：Data Analysis Methods:
The BER for the DMT signal was measured off-line. The comparison of complexity and BER for the four classification methods was analyzed.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
