Time attention analysis method for vibration pattern recognition of distributed optic fiber sensor

The distributed optical fiber sensing technology based on phase-sensitive optical time-domain reflectometer (Φ-OTDR) has the advantage of multi-point vibration monitoring simultaneously. To improve the robustness of the recognition system, a deep learning model is designed to extract time-frequency sequence correlation from signals and spectrograms. This lightweight plug-and-play convolutional neural network architecture achieves better performance with a lower number of extra parameters and computations. This study is verified on a vibration dataset containing eight different scenarios collected by a Φ-OTDR system, and the classification accuracy of 96.02% is achieved. [Display omitted] •A study of the global and local correlation of signal data in time frequency domain.•A targeted time attention model is designed to extract characteristics vibration signal feature.•A large scale of dataset of vibration scenes is collected and labeled.