Optical semantic communication through multimode fiber: from symbol transmission to sentiment analysis

Area of Science:

• Optical communication
• Information theory
• Signal processing

Background:

• Multimode fiber (MMF) communication systems face limitations in capacity and robustness.
• Conventional encoding schemes struggle to maximize spectral efficiency.
• Semantic communication offers a potential pathway to enhance data transmission.

Purpose of the Study:

• To propose and validate a novel optical semantic transmission scheme utilizing multimode fiber.
• To enhance data capacity and spectral efficiency in optical communication.
• To demonstrate the system's applicability in semantic analysis and improve noise tolerance.

Main Methods:

• Leveraging the frequency sensitivity of intermodal dispersion in MMFs for high-dimensional semantic encoding.
• Mapping symbols to 128 distinct frequencies at 600 kHz intervals.
• Implementing 4-level pulse amplitude modulation (PAM-4) for enhanced spectral efficiency.

Main Results:

• Achieved a seven-fold increase in capacity compared to conventional methods.
• Reached a spectral efficiency of 9.12 bits/s/Hz without decoding errors using PAM-4.
• Demonstrated accurate sentiment analysis on the IMDb dataset by encoding semantically similar symbols to adjacent frequencies.

Conclusions:

• The proposed MMF-based semantic communication scheme significantly enhances capacity and robustness.
• The system shows promise for applications in bandwidth-constrained and noisy optical communication environments.
• Semantic encoding improves noise tolerance, enabling effective data processing tasks like sentiment analysis.