Modulation Format Recognition Based on Coordinate Transformation and Combination in VLC System

With the development of modern communication technology, the new generation of communication network should have the characteristics of faster speed, higher bandwidth utilization, and wider coverage of users and business needs. At the same time, the intelligent optical network is also facing the challenges from the rapid growth of bandwidth demand and the rapid increase of different users services. Therefore, it is necessary to mix different modulation formats in the communication network to meet these requirements. Modulation format recognition (MFR) technology is therefore widely used at the receiving end to improve the performance of subsequent digital signal processing (DSP) algorithms. Compared with traditional machine learning (ML), deep learning (DL) has achieved amazing achievements in classification tasks in recent years. Therefore, we propose a coordinate merging algorithm (CMA) to process the data for DL. Compared with the traditional method of directly processing the in-phase and quadrature (IQ) components of the received signal, and another method we tried to fuse the weights of classification results of signals with different coordinates, the CMA method achieved higher accuracy and better performance against low signal-to-noise (SNR) ratios in visible light communication (VLC) system.