Recognition of orbital angular momentum modes based on deep learning and multiaperture speckle

Abstract

We propose a method for orbital angular momentum (OAM) mode recognition based on deep learning (DL) and multiaperture speckle (MS). This method works robustly in complex environments, such as angular disturbances and random obstacles. It also has a high OAM mode recognition accuracy (RA) at 99.5 %. RA remains high even at various propagation distances. Results show that the combination of the photon OAM and the number of the apertures can be used as a new dimension for structured light information transmission. The multiaperture optical beams can be used to construct a transmission channel with multiplexed information. As an example, an image has been transmitted in an experiment utilizing the multiaperture dimension of light. The bit error rate (BER) for image transmission is as low as 0.065 %. This work provides a new application scheme for free space optical communication.