Anti-Nonlinear Support Vector Machine Based Geometrically Shaping Photonic Terahertz Communication System

Aiming at the problems of low spectral efficiency and high transmission power in current photonic terahertz (THz) communication systems, a two channel signal modulation scheme based on independent sidebands (ISB) and geometric shaping (GS) is proposed, and the support vector machine (SVM) supervised learning algorithm is applied to the photonic THz communication system to achieve classification and judgment of received signal constellation points, thereby solving the problem of nonlinear effects. The simulation results show that under the transmission conditions of standard single-mode fiber (SSMF) with a carrier frequency of 90 GHz and a fiber transmission distance of 50 km, GS of the signal can achieve the best bit error rate (BER) when using a square 16QAM. At the same time, the error performance of the squared 16QAM signal with / without SVM algorithm was also measured, and the results showed that the algorithm can achieve a large received optical power (ROP) gain at a threshold of 3.8×10
^-3
.