OAM optical fiber transmission with reduced bit error rate based on turbo code encoding/decoding and deep learning demodulation

Deep learning-assisted OAM optical fiber transmission has become a hotspot because of its end-to-end and high-accurate demodulation characteristics. Recent studies report that nearly 100% recognition rate of demodulation can be achieved, but the actual bit error rate (BER) of data transmission is not very low. In order to maintain a high-accurate demodulation with high-fidelity transmission performance, we propose a feasible scheme that combines turbo code encoding/decoding and deep learning demodulation for OAM optical fiber communications. In the scheme, a turbo code encoder/decoder is designed as a bit error corrector and a pre-trained deep learning model is induced as an OAM mode demodulator, which ensures the high-accurate demodulation performance of OAM and the high-fidelity data transmission with reduced BER of the system. To verify the proposed scheme, we experimentally built an OAM optical fiber transmission system and transmitted a 256-grayscale image based on the system. A recognition rate of 99.69% and a transmission BER of 1.32 × 10^–3 demonstrate the excellent performance. In addition, better transmission performance can be achieved when combined with an improved OAM mapping strategy. Our scheme can be a new perspective for high-performance OAM fibre transmission.