Deep-learning-based high-resolution recognition of fractional-spatial-mode-encoded data for free-space optical...

Area of Science:

• Optics and Photonics
• Data Communication
• Artificial Intelligence

Background:

• Optical communication faces limitations in capacity improvement due to integer quantization of spatial degrees of freedom (DoF).
• Existing methods struggle to effectively increase data transmission rates using structured light.

Purpose of the Study:

• To develop a novel data transmission system utilizing fractional mode encoding and deep-learning decoding.
• To overcome the limitations of integer quantization in spatial DoF for enhanced communication capacity.

Main Methods:

• Employing spatial modes of Bessel-Gaussian beams separated by fractional intervals for data representation.
• Utilizing phase holograms for data encoding and a deep-learning classifier for decoding based on intensity profiles.
• Training a deep-learning model to recognize independent DoF and fractional mode differences.

Main Results:

• Successfully decoded data using a deep-learning classifier that only requires intensity profiles.
• Demonstrated simultaneous recognition of two independent DoF without a mode sorter.
• Achieved high-fidelity image transmission despite densely packed fractional modes.

Conclusions:

• Fractional mode encoding with deep-learning decoding offers a new approach for higher data rates in optical communication.
• The proposed scheme effectively utilizes spatial DoF for advanced optical communication systems.
• This research paves the way for next-generation high-capacity optical networks.