公开(公告)号：US11562224B2

公开(公告)日：2023-01-24

申请号：US16057828

申请日：2018-08-08

Applicant: University of Electronic Science and Technology of China

Inventor： Huijuan Wu ,  Jiping Chen ,  Xiangrong Liu ,  Yao Xiao ,  Mengjiao Wang ,  Bo Tang ,  Mingru Yang ,  Haoyu Qiu ,  Yunjiang Rao

IPC: G06N3/08 ,  G06F17/18 ,  G06N3/04 ,  G06N5/00 ,  G06N20/10 ,  G06N20/20

Abstract: A 1D-CNN-based ((one-dimensional convolutional neural network)-based) distributed optical fiber sensing signal feature learning and classification method is provided, which solves a problem that an existing distributed optical fiber sensing system has poor adaptive ability to a complex and changing environment and consumes time and effort due to adoption of manually extracted distinguishable event features. The method includes steps of: segmenting time sequences of distributed optical fiber sensing acoustic and vibration signals acquired at all spatial points, and building a typical event signal dataset; constructing a 1D-CNN model, conducting iterative update training of the network through typical event signals in a training dataset to obtain optimal network parameters, and learning and extracting 1D-CNN distinguishable features of different types of events through an optimal network to obtain typical event signal feature sets; and after training different types of classifiers through the typical event signal feature sets, screening out an optimal classifier.

公开(公告)号：US20180357542A1

公开(公告)日：2018-12-13

申请号：US16057828

申请日：2018-08-08

Applicant: University of Electronic Science and Technology of China

Inventor： Huijuan Wu ,  Jiping Chen ,  Xiangrong Liu ,  Yao Xiao ,  Mengjiao Wang ,  Bo Tang ,  Mingru Yang ,  Haoyu Qiu ,  Yunjiang Rao

IPC: G06N3/08 ,  G06N3/04 ,  G06F17/18

CPC classification number: G06N3/08 ,  G06F17/18 ,  G06N3/04

Abstract: A 1D-CNN-based ((one-dimensional convolutional neural network)-based) distributed optical fiber sensing signal feature learning and classification method is provided, which solves a problem that an existing distributed optical fiber sensing system has poor adaptive ability to a complex and changing environment and consumes time and effort due to adoption of manually extracted distinguishable event features, The method includes steps of: segmenting time sequences of distributed optical fiber sensing acoustic and vibration signals acquired at all spatial points, and building a typical event signal dataset; constructing a 1D-CNN model, conducting iterative update training of the network through typical event signals in a training dataset to obtain optimal network parameters, and learning and extracting 1D-CNN distinguishable features of different types of events through an optimal network to obtain typical event signal feature sets; and after training different types of classifiers through the typical event signal feature sets, screening out an optimal classifier.

公开(公告)号：US20180080812A1

公开(公告)日：2018-03-22

申请号：US15812996

申请日：2017-11-14

Applicant: University of Electronic Science and Technology of China

Inventor： Huijuan Wu ,  Wei Zhang ,  Xiangrong Liu ,  Yunjiang Rao

IPC: G01H9/00 ,  G01V1/18 ,  G01V1/28

CPC classification number: G01H9/004 ,  G01V1/18 ,  G01V1/208 ,  G01V1/226 ,  G01V1/288 ,  G06F17/12 ,  G06N5/04 ,  G06N7/005 ,  H04B10/25

Abstract: A distributed optical fiber sensing signal processing method for safety monitoring of underground pipe network, which belongs to infrastructure safety monitoring field, which is aimed to improve the intelligent ability of detection and identification of the existing distributed optical fiber sound/vibration sensing system under complex application conditions. The present invention utilizes the distributed optical fiber sound/vibration sensing system to pick up the sound or vibration signal of the whole line along the detection cable; and the customized short time feature and long time feature are respectively extracted from the relative quantity of the sound or the vibration signal at each spatial point in the whole monitoring range. The Bayesian identification and classification network at each spatial point is constructed and trained based on the prior knowledge of the collected signal features and their different background noises.