Machine learning approach to OAM beam demultiplexing via convolutional neural networks

Area of Science:

• Optical communication systems
• Free-space optical communication
• Information theory

Background:

• Orbital angular momentum (OAM) beams enhance channel capacity in free-space optical communication.
• Traditional demultiplexing relies on orthogonality and often requires precise alignment and specialized hardware.
• Atmospheric turbulence degrades signal quality and complicates demultiplexing.

Purpose of the Study:

• To propose and evaluate a novel convolutional neural network (CNN)-based technique for demultiplexing OAM beams.
• To demonstrate the simplicity and robustness of the CNN method compared to traditional approaches.
• To assess the CNN method's performance under various conditions, including atmospheric turbulence and sensor noise.

Main Methods:

• Capturing images of multiplexed OAM beam intensity patterns.
• Training a convolutional neural network (CNN) as a classifier to demultiplex OAM modes.
• Comparing the CNN method against conjugate mode sorting under simulated atmospheric turbulence.
• Evaluating the CNN method's resilience to sensor noise, photon detection levels, pixel count, and training set size.

Main Results:

• The CNN-based demultiplexing method achieved >99% accuracy for demultiplexing combinatorially multiplexed OAM modes.
• The CNN method significantly outperformed traditional conjugate mode sorting, especially under high turbulence.
• The technique demonstrated robustness against sensor noise, varying photon detection counts, pixel numbers, unknown turbulence levels, and training set sizes.

Conclusions:

• CNN-based demultiplexing offers a simple, hardware-efficient alternative for OAM beam separation in free-space optical communication.
• This AI-driven approach overcomes limitations of traditional methods, including alignment sensitivity and strict orthogonality requirements.
• The proposed method is highly accurate and robust, showing significant promise for practical OAM communication systems.