A Fast Accurate Attention-Enhanced ResNet Model for Fiber-Optic Distributed Acoustic Sensor (DAS) Signal Recognition in Complicated Urban Environments

Abstract

The fiber-optic distributed acoustic sensor (DAS), which utilizes existing communication cables as its sensing media, plays an important role in urban infrastructure monitoring and natural disaster prediction. In the face of a wide, dynamic environment in urban areas, a fast, accurate DAS signal recognition method is proposed with an end-to-end attention-enhanced ResNet model. In preprocessing, an objective evaluation method is used to compare the distinguishability of different input features with the Euclidean distance between the posterior probabilities classified correctly and incorrectly; then, an end-to-end ResNet is optimized with the chosen time-frequency feature as input, and a convolutional block attention module (CBAM) is added, which can quickly focus on key information from different channels and specific signal structures and improves the system recognition performance further. The results show that the proposed ResNet+CBAM model has the best performance in recognition accuracy, convergence rate, generalization capability, and computational efficiency compared with 1-D CNN, 2-D CNN, ResNet, and 2-D CNN+CBAM. An average accuracy of above 99.014% can be achieved in field testing; while dealing with multi-scenario scenes and inconsistent laying or burying environments, it can still be kept above 91.08%. The time cost is only 3.3 ms for each signal sample, which is quite applicable in online long-distance distributed monitoring applications.

Quantitative Analysis of φ-OTDR Spatial Resolution Influenced by NLM Parameters

“…Lastly, the NLM parameter that has the greatest effect on signal broadening is determined, and the extent of signal broadening is evaluated when the NLM parameters are optimal. This work helps to improve the SR and MA of ϕ-OTDR and the performance of vibration signal recognition [29].…”

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

“…In order to obtain the annotations, it is necessary to use the existing model to force the alignment between the training data sequence and the annotated sequence, which is time consuming and often has biases in preparing these annotations. LSTMs [30], Bi-LSTMs [31], or CNNs [32] that incorporate an attention mechanism can, without forcing alignment, make the features extracted by the model focus more on locally valid details of the vibration signal rather than on global features, which helps to improve the overall recognition rate.…”