A multi-regressor location method based on a semiconductor optical amplifier (SOA) ring cavity reservoir computing (RC) is proposed for the distributed optical fiber sensing system implemented with time-delay RC. Multiple linear regressors are integrated into the RC output layer. One provides a global location across the entire sensing fiber, and the others correspond to the segmented location. The regressors are trained via ridge regression in various sensing ranges. Disturbance location is achieved by first thresholding the segment location results and then selecting the result closest to the global location result. The experimental results show that the proposed method enables any position location without collecting the disturbance sensing signals at all positions, demonstrates superiority over the conventional time difference method in location accuracy, and significantly reduces training requirements compared with classification-based RC methods.
Liu et al. (Tue,) studied this question.