Spatial pilot-aided fast-adapted framework for stable image transmission over long multi-mode fiber.

Area of Science:

• Optical Communications
• Signal Processing
• Machine Learning

Background:

• Multi-mode fiber (MMF) offers high capacity for spatial information transmission.
• MMF's scattering and time-varying characteristics challenge stable, long-term data transmission.
• Existing methods struggle with the dynamic nature of MMF channels.

Purpose of the Study:

• To develop a robust framework for stable spatial image transmission over unstable MMF.
• To address the challenges posed by MMF channel variations for practical applications.
• To enable accurate and continuous data transmission in complex MMF environments.

Main Methods:

• A spatial pilot-aided learning framework using reference image frames.
• A fast-adapt network training scheme for online model updates.
• Two pilot-insertion strategies evaluated for various transmission scenarios.

Main Results:

• Achieved transmission accuracy exceeding 92% over hours on 100m unstable MMFs.
• Pilot frame overhead was approximately 2% of the total data.
• Fast-adapt learning trained <2% of network parameters, reducing computation time by 70%.

Conclusions:

• The proposed framework enables stable and accurate spatial transmission over MMF, overcoming channel instability.
• The fast-adapt learning approach significantly improves efficiency and reduces computational load.
• This technology enhances the feasibility of MMF for practical, complex spatial transmission applications.