Modulation Format Identification Based on Signal Constellation Diagrams and Support Vector Machine

In coherent optical communication systems, where multiple modulation formats are mixed and variable, the correct identification of signal provides foundation for subsequent performance improvement using digital algorithms. A format (MFI) scheme based on constellation diagrams support vector machine (SVM) is proposed. Firstly, divided by fractal dimension weighted linear least squares (WLS-FD) algorithm, (FD) in each region calculated, which regarded as one image features. Then, feature values different directions extracted gray-level co-occurrence matrix (GLCM), their mean variance another feature. Finally, two features input into classifier constructed SVM to achieve MFI systems. To verify feasibility superiority scheme, we compare it with higher-order statistical (HOS) features, GLCM FD respectively. Further, built a 30 GBaud system fiber lengths 80 km 120 km, signal-to-noise ratio (OSNR) ranges from 0 dB dB. The proposed identifies seven formats: QPSK, 8QAM, 16QAM, 32QAM, 64QAM, 128QAM, 256QAM. results show that compared other three schemes, our has better accuracy at low OSNR. addition, this can reach 100% when OSNR ≥ 10