Convolutional neural network for self-mixing interferometric displacement sensing.

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Laser Technology

Background:

• Self-mixing interferometry is a robust technique but signal interpretation challenges limit its practical use.
• Complex self-mixing signals hinder accurate displacement measurements in semiconductor laser systems.
• Developing advanced signal processing methods is crucial for improving interferometric measurement availability.

Purpose of the Study:

• To develop a novel method for reconstructing target displacement from self-mixing signals using artificial intelligence.
• To overcome the practical difficulties in interpreting complex self-mixing interferometry signals.
• To enhance the reliability and applicability of semiconductor laser-based self-mixing measurement systems.

Main Methods:

• A convolutional neural network (CNN) was employed to analyze and reconstruct displacement data.
• The CNN was trained using periodic displacement patterns generated from a semiconductor laser.
• The model was tested under various alignment conditions and with different experimental setups.

Main Results:

• The trained CNN successfully reconstructed arbitrarily complex target displacements from self-mixing signals.
• The method demonstrated robustness across different alignment conditions and setups.
• The AI-based approach significantly improved the interpretability and reliability of self-mixing measurements.

Conclusions:

• Convolutional neural networks offer a powerful solution for interpreting complex self-mixing signals in semiconductor lasers.
• This AI-driven approach enhances measurement accuracy and broadens the applicability of self-mixing interferometry.
• The validated method shows potential for generalization to modulated schemes and other self-mixing sensing tasks.