Using Phase-Sensitive Optical Time Domain Reflectometers to Develop an Alignment-Free End-to-End Multitarget Recognition Model

This pattern recognition method can effectively identify vibration signals collected by a phase-sensitive optical time-domain reflectometer (Φ-OTDR) and improve the accuracy of alarms. An alignment-free end-to-end multi-vibration event detection based on Φ-OTDR is proposed, detecting different events in frequency bands. The pulse accumulation cancellers determine location events. local differential demodulates variation signals. After extraction signal time-frequency features sliding window, convolution neural network (CNN) further extracts features. It analyzes temporal relationship each group using bidirectional long short-term memory (Bi-LSTM). Finally, connectionist classification (CTC) used to label unsegmented sequence data achieve single multiple targets. Experiments show that this process 8563 data, containing 5 bands acoustic sensing signal, system F1 score 99.49% with time 2.2 ms. highest response 1 kHz. available quickly efficiently when demodulated contains