Real-time demodulation of complex FBG spectra via dictionary-enhanced sparse Bayesian learning

Area of Science:

• Optical Engineering
• Signal Processing
• Machine Learning

Background:

• Fiber Bragg gratings (FBGs) are crucial for optical sensing.
• Demodulating complex FBG spectra with overlapping peaks is challenging.

Purpose of the Study:

• To propose a novel dictionary Gaussian sparse Bayesian learning (DG-SBL) framework.
• To enable comprehensive and accurate demodulation of complex FBG spectra.

Main Methods:

• A two-stage approach: dictionary learning and real-time demodulation.
• Alternating iterative optimization for dictionary learning with K-sparsity constraint.
• Hybrid tracking strategy utilizing direct matching and covariance-free Bayesian algorithm for reconstruction.

Main Results:

• Achieved >99.99% cosine similarity in waveform reconstruction.
• Demonstrated fast processing: 0.021 s/frame (learning), <0.2 ms latency (tracking).
• Directional weighting effectively separated overlapping signals.

Conclusions:

• The DG-SBL framework provides accurate and efficient FBG spectra demodulation.
• The method is scalable for real-time sparse recovery in industrial monitoring.
• Significantly reduces computational complexity for waveform reconstruction.