Neural network-assisted signal processing in Brillouin optical correlation-domain sensing for potential high-speed...

Area of Science:

• Optoelectronics
• Signal Processing
• Machine Learning

Background:

• Brillouin-based distributed sensing traditionally extracts Brillouin frequency shift (BFS).
• Conventional methods often involve complex signal processing, limiting real-time applications.
• Neural networks (NNs) offer a potential alternative for BFS extraction.

Purpose of the Study:

• To apply NNs with adaptive BFS incremental steps to Brillouin optical correlation-domain sensing.
• To improve the accuracy and operational speed of BFS extraction.
• To provide a solution for online signal processing in real-time Brillouin sensing.

Main Methods:

• Utilized neural networks (NNs) with adaptive BFS incremental steps for signal processing.
• Applied the method to Brillouin optical correlation-domain sensing.
• Compared NN performance against conventional curve fitting methods.

Main Results:

• NNs improved BFS measurement accuracy by 1.6-2.7 times in experiments.
• NNs accelerated BFS extraction speed by up to 5000 times in experiments.
• Simulated signals showed 2-3 times accuracy improvement and 1000 times speedup.

Conclusions:

• NNs with adaptive BFS incremental steps offer a significant advancement in Brillouin sensing.
• The proposed method enhances both accuracy and speed for BFS extraction.
• This approach is a viable solution for real-time signal processing in Brillouin sensing systems.