Distributed Optical Fiber Sensing Signal Recognition Based on Deep Residual Shrinkage Network

Deep residual shrinkage network (DRSN) is proposed to enhance the recognition effect of distributed optical fiber sensing signals. The experimental data are collected by distributed optical fiber sensing perimeter security system in which six kinds of vibration signals are studied, including slapping, walking, rolling, noise, climbing and raining. The experimental results show that the recognition accuracy of DRSN method is higher than that of CNN and ResNet methods. The average test accuracy of DRSN method is up to 98.017%. Compared with ResNet method, the average test accuracy of DRSN method is improved by about 1.1 %. Compared with CNN method, the average test accuracy of DRSN method is improved by about 2.5%. Moreover, DRSN method is obviously superior to CNN and ResNet methods according to the evaluation indexes of precision, recall and F1-score.