A deep learning approach for trustworthy high-fidelity computational holographic orbital angular momentum communication

Orbital angular momentum (OAM) holography is becoming a promising technology for image encryption, optical transmission, and storage because of its excellent fidelity, orthogonality, and security. Benefiting from the powerful ability of machine learning to learn from big data features, a computational holographic orbital angular momentum (OAM) communication method using OAM hologram encoding and machine learning decoding is proposed. The OAM information representing the grayscale of the images is encoded into different holograms. Subsequently, using a well-trained convolutional neural network, the holograms carrying arbitrary image information can be accurately transmitted and translated, and the hidden OAM information is readout quickly and accurately as an added confidential channel. Topological charge digits can be arranged to form grayscale images or serial codes. Such a computational holographic OAM communication method can be used for extended channels with high security and complexity. In addition, this method can be applied in areas of confidential digital modulation/demodulation and encrypted communication, as well as expand the transmission capacity.