Estimation of variable optical feedback coupling factor for self-mixing interferometry by signal-to-image translation

Accurate estimation of the optical feedback coupling factor <i>C</i> is essential for the reliable operation of laser-based self-mixing interferometry (SMI) sensors, particularly in variable feedback conditions. In this work, we present a novel deep learning-based method, to the best of our knowledge, to estimate the time-varying <i>C</i> factor from SMI signals under strong, moderate, and weak feedback conditions. The proposed approach leverages a transformation of one-dimensional (1D) SMI signals into two-dimensional (2D) image representations, allowing convolutional neural networks to extract context-rich features that enhance estimation performance. Unlike traditional image-processing tasks, this method formulates the problem as a signal-to-image translation, tailored for sensor parameter inference. Experimental results demonstrate that this 2D-based approach outperforms state-of-the-art recurrent neural network models, including LSTM and transformer architectures, particularly in terms of robustness to changes in sampling frequency, displacement amplitude, and feedback regime. The methodology is generic and applicable to a wide range of sensor signal analysis tasks. To support reproducibility and practical adoption, we provide a publicly available implementation of our approach.