Encrypted information transmission via high-resolution elliptical vortex modes based on deep learning

The elliptical vortex beam exhibits two degrees of freedom, namely topological charge and ellipticity, both of which can greatly improve the information transmission capacity and processing capability for optical communication. Accurate detection of the two degrees of freedom in the elliptical vortex modes by the receiver is key to determining the communication capability. This paper presents an improved ResNet architecture built upon the convolutional neural network method, to accurately identify the encrypted dual-mode superimposed high-resolution interference patterns of the elliptical vortex beam. Our results show that even with a topological charge resolution of 0.01 and an ellipticity resolution of 0.1, the recognition accuracy for the two degrees of freedom is 88.12% and 99.85%, respectively. Then, the elliptical vortex beam multiplexing encryption system is used to transmit the cat image, yielding the similarly favorable transmission outcomes. At a topological charge resolution of 0.1, the cat image transmission accuracy can reach 99.64%. Such results provide new perspective for the manipulation and exploitation of the elliptical vortex beam, and have significant implications for free optical communication based on orbital angular momentum.