Laser self-mixing interferometry for direct displacement reconstruction using deep learning

Abstract

In this paper, we propose a displacement reconstruction method based on laser self-mixing interferometer using deep learning. A novel Direct Displacement Reconstruction Network (DDR-Net) has been designed to reconstruct displacement from self-mixing interference (SMI) signals. The DDR-Net takes the original SMI signal as the input, using the target displacement as the training label to achieve an end-to-end regression task. Experimental results demonstrate that DDR-Net effectively reconstructs target displacement using training data collected from real experimental environments, showcasing robust generalization across different displacement modes and noisy conditions. Multiple random experiments reveal that the relative measurement error of the reconstructed displacement consistently remains below 1 %. This method significantly simplifies the data processing workflow, highlighting its potential for applications in laser self-mixing sensing and offering new insights for tackling complex measurement tasks.