Optical fiber distributed acoustic sensing signal identification method, system, device and medium
By combining transfer learning and semi-supervised learning methods, and utilizing the ResNet-34 network and the improved FixMatch algorithm, the problems of transferability and unlabeled sample utilization of fiber optic distributed acoustic sensors in cross-scene recognition were solved, achieving higher recognition accuracy and faster model adaptation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2023-12-06
- Publication Date
- 2026-07-21
AI Technical Summary
Existing fiber optic distributed acoustic sensor (DAS) signal recognition models lack transferability when recognizing across scenes, and semi-supervised learning methods fail to make full use of unlabeled samples, resulting in poor recognition performance.
By combining transfer learning and semi-supervised learning, a signal recognition model is constructed using the ResNet-34 network. Knowledge distillation and adaptive knowledge consistency (AKC) are used for transfer learning. The feature distribution constraints of labeled and unlabeled samples are utilized, and the improved FixMatch algorithm is combined for semi-supervised training to enhance the model's adaptability in cross-scene recognition.
It improves the accuracy of cross-scene recognition and the ability to quickly deploy models, and can effectively utilize source domain knowledge and labeled or unlabeled samples in the target domain to achieve higher recognition accuracy and faster model convergence.
Smart Images

Figure CN117636871B_ABST