Abstract
To improve the information transmission ability of point-to-point optical fiber communication systems, we propose and experimentally demonstrate an optical fiber communication system based on intelligent joint source-channel coded modulation (JSCCM-OFC). Instead of encoding information into bits by source coding and employing channel coding, the JSCCM-OFC system codes and modulates the information source into discrete-time analog symbols jointly using deep learning. The generated discrete-time analog symbols are then directly transmitted through an optical fiber. We designed language attention and dual-attention residual networks as the JSCCM for text and image transmission, respectively. For interaction between the information layer and the optical physical layer, we incorporated a convolutional neural network into the joint demodulation and decoding (JDD) network and implemented joint optimization in the receiver. Compared with the bit-based structure, the JSCCM-OFC system achieved higher information compression and obtained a more stable performance, especially in the low-received optical power regime. Moreover, JSCCM enhanced the robustness of the system against optical link impairments.
PDF Article
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription