Distributed Optical Fiber Sensing System To Trigger Pattern Recognition Research

Fiber distributed sensing technology has nearly 20 years of development history, which uses a few-kilometer-long fiber as a unit of sensor and signal transmission channel, and the outside world can be measured along the fiber physical space with timing information. In practical application projects, it is applicable to long-distance and big regional scope of monitoring and security control.Previous research on distributed fiber sensing system focused on improving the structure of the optical system and improving the algorithm, but for the system used in the actual monitoring projects, signal pattern recognition is not researched. This article proposed a multi-scale dynamic decision-making DTW algorithm to provide an effective method for pattern recognition applications in fiber distributed sensing system, which contributed practical significance to fiber distributed sensing system applications.After briefly introducing the distributed fiber sensing technology and pattern recognition as the background, this paper introduced the status and significance of research on pattern recognition in distributed fiber sensing system. Then, on the signal analysis of characteristics in the system, the signal pattern in the system is summed up as well as specific issues need to be solved, including: the cyclical signal spectrum destruction, the time-domain signal alignment, and the characteristics of the template choice. Against the system-specific problems, a dynamic multi-scale decision-making DTW algorithm is introduced for pattern recognition and its principles and methods is proved. The algorithm could pick-up parameters in signal's frequency spectrum in accordance with sub-scale segment, and according to the location of the destruction of the spectrum, dynamic adjustment of the decision-making weight is made to match the further pattern recognition process. Finally, a set of data from a practical system is collect to verify that the algorithm for pattern recognition in the distributed fiber sensing system is effective.