Binary signaling design for visible light communication: a deep learning framework.

Area of Science:

• Electrical Engineering
• Computer Science
• Optical Communications

Background:

• Visible Light Communication (VLC) systems offer a promising alternative for wireless data transmission.
• Achieving dimming support in VLC systems using On-Off Keying (OOK) presents a combinatorial design challenge for constant weight codes (CWC) over signal-dependent noise channels.

Purpose of the Study:

• To develop a deep learning framework for designing OOK-based binary signaling transceivers in dimmable VLC systems.
• To address the combinatorial design problem of CWCs for dimmable OOK signals.

Main Methods:

• An autoencoder (AE) framework was employed to learn an encoder-decoder neural network.
• Optical channel layers and binarization techniques were integrated to model the physical and discrete nature of OOK-based VLC systems.
• End-to-end training was utilized to design and optimize the VLC transceiver.

Main Results:

• The proposed deep learning framework successfully designed an OOK-based binary signaling transceiver for dimmable VLC.
• Numerical results demonstrated superior performance of the proposed transceiver compared to baseline CWC schemes.
• The AE approach effectively handled the signal-dependent noise channels and combinatorial design problem.

Conclusions:

• Deep learning, specifically the autoencoder approach, provides an effective framework for designing advanced VLC transceivers.
• The proposed method offers a significant performance improvement over existing CWC schemes for dimmable OOK-based VLC systems.
• This work paves the way for more efficient and robust visible light communication systems.