Modulation format recognition in a UVLC system based on reservoir computing with coordinate transformation and...

Area of Science:

• Optical Communications
• Signal Processing
• Machine Learning

Background:

• Modulation format recognition (MFR) is crucial for adaptive optical systems.
• Deep learning-based MFR algorithms show success but are computationally intensive, especially in complex underwater visible light communication (UVLC) channels.
• Complex neural network (NN) structures hinder real-time processing and efficient allocation in UVLC.

Purpose of the Study:

• To propose a lightweight and efficient MFR method for UVLC systems.
• To reduce the computational complexity and improve the processing speed of MFR.
• To investigate the performance of reservoir computing (RC) for MFR tasks in UVLC.

Main Methods:

• Developed a reservoir computing (RC) based method with significantly fewer trainable parameters (0.3% of NN methods).
• Implemented advanced feature extraction algorithms, including coordinate transformation and folding algorithms, to enhance RC performance.
• Tested the RC method on six modulation formats: OOK, 4QAM, 8QAM-DIA, 8QAM-CIR, 16APSK, and 16QAM.

Main Results:

• The proposed RC-based MFR method achieved training times in seconds.
• Recognition accuracy exceeded 90% for most modulation formats, reaching close to 100% under varying LED pin voltages.
• The method demonstrated high efficiency and accuracy for MFR in UVLC.

Conclusions:

• Reservoir computing offers a computationally efficient and accurate alternative for MFR in UVLC.
• Feature extraction techniques like coordinate transformation and folding are effective in improving RC performance for MFR.
• The study provides guidance for designing effective RC systems that balance accuracy and computational cost for MFR applications.