Deep-learning-based recognition of composite vortex beams through long-distance and moderate-to-strong atmospheric turbulence
Siwen Cai, Zhihui Li, Zheqiang Zhong, Bin Zhang
Abstract
Orbital angular momentum (OAM), as a physical dimension of light, has been demonstrated to enhance the channel capacity and turbulence resistance of free-space optical (FSO) communication. However, the channel crosstalk in OAM-based FSO communication inevitably increases with transmission distance and turbulence intensity. Here, we propose a deep-learning-based recognition of a composite vortex beam to extend the regime of moderate-to-strong turbulence and long-distance FSO links. The composite vortex beam is generated by a coherent combination of two subbeams carrying different helical charges and phase delays, providing its helical charges and phase delay as new multiplexing dimensions and exhibiting better turbulence resistance compared to a single subbeam. We also developed a modified regular network to achieve the high-accuracy recognition of a composite vortex beam over a long distance at moderate-to-strong atmospheric turbulence. We believe that our approach has potential in deep-learning-based OAM high-capacity communication systems.