Estimation of Displacement Direction Based on Self-mixing Interferometry and Convolution Neural Network

Self-mixing interferometry (SMI) is a promising interferometric measurement technology with unique system structure. It has an advantage that conventional two-beam interferometry do not have, i.e., movement directions of the target to be measured can be determined by a single-channel interferometric signal due to the existence of linewidth enhancement factor in lasers. However, movement directions are difficult to be determined when the optical feedback is weak. In this work, an algorithm is proposed for determination of SMI-based displacement directions based on convolution neural network (CNN). We used Python language and the third-party libraries NumPy to complete numerical calculation as well as TensorFlow to establish the CNN. The simulation results shows that displacement directions are able to be determined with the accuracy higher than 94.8% when the optical feedback factor is low to 0.1.