CN110135283A - The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm - Google Patents

The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm Download PDF

Info

Publication number
CN110135283A
CN110135283A CN201910336913.3A CN201910336913A CN110135283A CN 110135283 A CN110135283 A CN 110135283A CN 201910336913 A CN201910336913 A CN 201910336913A CN 110135283 A CN110135283 A CN 110135283A
Authority
CN
China
Prior art keywords
signal
optical fiber
test sample
defence system
fiber perimeter
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201910336913.3A
Other languages
Chinese (zh)
Inventor
方捻
王宁宁
王陆唐
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN201910336913.3A priority Critical patent/CN110135283A/en
Publication of CN110135283A publication Critical patent/CN110135283A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/22—Source localisation; Inverse modelling

Landscapes

  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention discloses a kind of signal recognition methods of optical fiber perimeter defence system based on FastDTW algorithm.The acquisition different types of output signal of optical fiber perimeter defence system in advance；Collected signal is divided into several signal segments according to classification；Each type of type signal segment is selected, it is normalized, is marked as reference template, and to each template；Freshly harvested signal to be identified is similarly divided and normalized, as test sample；Each test sample is calculated at a distance from the best regular path of all reference templates using FastDTW algorithm；Finally, being chosen and signal classification of label of each test sample apart from shortest reference template as the test sample according to Nearest neighbor rule.The present invention is not necessarily to extract the feature of signal, and identification process is simple and easily realizes；Higher correct recognition rata can be reached without excessive training sample set, the less situation of this training sample of the signal identification suitable for optical fiber perimeter defence system.

Description

The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm
Technical field
The present invention relates to a kind of signal recognition methods, especially a kind of to be based on FastDTW (Fast Dynamic Time Warping, quick dynamic time warping) algorithm optical fiber perimeter defence system signal recognition method.
Background technique
In recent years, optical fiber perimeter defence system is commonly called as fiber fence, due to more dry than fence good reliability, anti-electromagnetism Ability is disturbed to have received widespread attention by force.If being laid with optical fiber with buried mode, system also has concealment, is not easy to be destroyed. However, rate of false alarm height is a key factor for influencing its performance at present.Various signal recognition methods are at analysis in recent years The detection signal of Ricoh's fibre circumferential protective system, it is intended to find out really invasion signal, reduce rate of false alarm.It is each for studying more The signal recognition method that kind is combined based on machine learning algorithm with feature extraction algorithm.But these methods in recognition accuracy and Good balance cannot be reached between recognition time, and need very big training set to train sorter model.In addition, these melt The identification process of conjunction scheme is generally all more complicated, and signal identification is made to lose timeliness, to affect its practical application.This Invention Announce it is a kind of it is very easy, efficiently and meet the recognition methods of practical application request.
Summary of the invention
It is an object of the invention to aiming at the defects existing in the prior art, provide a kind of optical fiber based on FastDTW algorithm The signal recognition method of circumferential protective system.
In order to achieve the above objectives, the present invention adopts the following technical scheme:
A kind of signal recognition method of the optical fiber perimeter defence system based on FastDTW algorithm, includes the following steps:
1) the different types of output signal of optical fiber perimeter defence system is acquired in advance；
2) collected signal is divided into several signal segments according to classification；
3) each type of type signal segment is selected, it is normalized, as reference template, and to each A template is marked；
4) freshly harvested signal to be identified is similarly divided and normalized, as test sample；
5) each test sample is calculated at a distance from the best regular path of all reference templates using FastDTW algorithm；
6) it according to Nearest neighbor rule, chooses with label of each test sample apart from shortest reference template as the test The signal classification of sample.
The working principle of the invention and feature:
Dynamic time warping (DTW) is a kind of algorithm of similitude for measuring two unequal time serieses of length, more Identification for voice signal.It stretches time series to be compared, referred to as cycle tests or compressed, until with ginseng The length for examining template is consistent, to facilitate the similitude for comparing the two.Specifically, it is by certain constraint condition to test specimens This progress is non-linear regular, i.e., the time shaft of test sample is non-linearly mapped to the time shaft of reference template, is tested The sum of the distance between all data points of sample and reference template, then obtaining one by Dynamic Programming makes between two sequences The shortest regular path of cumulative distance, to both achieve the purpose that best match.
It is respectively the reference template and test sample X, Y of M, N for length, has: X=[x1,x2,x3,…,xM], Y=[y1, y2,y3,…,yN].To be aligned the two samples, the grid matrix of a M*N is constructed, reaches lattice point (M, N) from lattice point (1,1) Path is regular path.Need to meet continuity, monotonicity and boundary condition for the selection in regular path, i.e., it cannot be across Some point go to match, the element in lattice point must be carried out with time dullness, regular path must from the beginning of time series to Terminal.By continuity and monotonicity it is found that the direction for reaching any one lattice point (i, j) only has 3.If xiAnd yjBetween two o'clock Distance be d (xi,yj), every lattice point, path distance gradually adds up, and the cumulative distance in regular path is denoted as Dist, reaches Cumulative distance at lattice point (m, n) isHere, the distance between two o'clock generallys use Europe Formula distance metric.It is found using dynamic programming method apart from shortest path, state transition equation are as follows:
Dist(xm,yn)=d (xm,yn)+min{Dist(xm-1,xn),Dist(xm-1,xn-1),Dist(xm,xn-1)}
FastDTW is the fast algorithm that proposes in order to reduce the Time & Space Complexity of DTW algorithm, is realized linear Space complexity.Firstly, coarsening time sequence is by being averaged to data point adjacent in time series to reduce the time The length or resolution ratio of sequence, repeatedly roughening generates the time series of multiple and different resolution ratio.With low resolution when Between find best regular path with the DTW algorithm of standard in sequence, the path is then projected into a higher resolution In time series, i.e., time series is refined, it is original twice that resolution ratio, which increases,.Therefore, regular in the case of low resolution The single mesh point that path is passed through will be at least mapped on 4 mesh points, then find local optimum path in view field.It throws The degree of refinement in shadow zone domain is determined that the value of r is generally 1 or 2, and the range of the bigger search of r is bigger, easier by extension radius r Find the regular path of global optimum.Grid matrix in FastDTW algorithm is filled only with the phase in the path under previous resolution ratio The length in neighbouring region, regular path linearly increases with the increase of the length of list entries, and the complexity of algorithm is reduced to O (N)。
Obtained by FastDTW algorithm the best regular path between each test sample and all reference templates away from From choosing classification of the label of reference template corresponding to the smallest distance as each test sample using Nearest neighbor rule Classification is to get the recognition result for arriving all signals.The mathematic(al) representation of Nearest neighbor rule is Ci=argmin (Dij), in formula, Ci It is the recognition result of i-th of test sample, DijIt is the best regular path between i-th of test sample and j-th of reference template Distance.
Finally, passing through correct recognition rata (IR), it may be assumed thatTo measure signal recognition method of the invention Accuracy, in formula, NcIt is the number of samples correctly identified, NtFor the number of all test samples.
The present invention compared with prior art, has following obvious prominent substantive distinguishing features and remarkable advantage:
The present invention is not necessarily to extract the feature of signal, and identification process is simple and easily realizes；When there is the invasion signal of new type When, the signal need to be only added in reference template, not need additional training；In addition, just without excessive training sample set It can achieve higher correct recognition rata, this training sample of signal identification for being highly suitable for optical fiber perimeter defence system is less The case where.
Detailed description of the invention
Fig. 1 is optical fiber perimeter defence system structural schematic diagram in embodiment 1.
Fig. 2 is the flow chart of the signal recognition method of the optical fiber perimeter defence system based on FastDTW algorithm.
Fig. 3 is the segment of one group of four kinds of type signal divided according to type.
Fig. 4 is the best regular path of first test sample and first reference template.
Specific embodiment
The preferred embodiment of the present invention is described with reference to the drawings as follows:
Embodiment 1
The distributed optical fiber sensing system as shown in Figure 1 based on linear type Sagnac interferometer is built in the lab to make For circumferential protective system, including wideband light source 1, circulator 2, coupler 3, Polarization Controller 4, delay optical fiber 5, coupler 6, biography Photosensitive fine 7, faraday rotator mirror 8, photodetector 9, data Collection & Processing System 10；The wideband light source 1 connects The port I of circulator 2, the port a of the port the II connection coupler 3 of the circulator 2, the port b of the coupler 3 connects coupling The port d of clutch 6, the port c of the coupler 3 connect the port e of coupler 6, institute by Polarization Controller 4, delay optical fiber 5 The port f for stating coupler 6 is connected by sensor fibre 7 with faraday rotator mirror 8, the port the III connection of the circulator 2 Photodetector 9, the photodetector 9, which changes into optical signal after electric signal, enters data acquisition and information processing system 10.
In the present embodiment, wideband light source 1 uses central wavelength for the ASE light source of 1550nm.Circulator 2 and coupler 3,6 It is the production of Shanghai Han Yu Fibre Optical Communication Technology Co., Ltd, the splitting ratio of coupler is 50:50.Polarization Controller 4 is using beauty The optical fiber squeezer (PLC-001) of General Photonics company, state.All optical fiber are all made of G.652 standard single-mode fiber. Faraday rotator mirror 8 is faraday's reflecting mirror that Shandong Zhaojin Group Co., Ltd. generates.Photodetector 9 flies for Shenzhen The PIN-TIA detector of logical company's production.Data Collection & Processing System 10 is by a conventional microcomputer and Britain PICO 5203 digital oscilloscope of PicoScope of company forms, oscillograph the data transmission of acquisition to computer, it is soft with Matlab Part programmed process obtains signal identification result.
The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm, as shown in Fig. 2, including following step It is rapid:
1) different types of output signal is acquired by the optical fiber perimeter defence system put up；
2) collected signal is divided into several signal segments according to type；
3) each type of type signal segment totally 20, including 5 signal segments without friction, 5 beats are chosen Signal segment, 5 walk signal segments and 5 tap segment twice, it is normalized respectively, as reference mould Plate, and 0,1,2 and 3 are respectively labeled as to the signal of above four seed type.Fig. 3 gives the segment of one group of four kinds of type signal, (1) respectively without friction, walking (2), beat signal (3) and knocking (4) twice.
4) freshly harvested signal to be identified is similarly divided and normalized, obtained test set includes 86 Signal segment, is 20 walk signal segments, 20 knocking segments twice respectively, 20 beat signal segments, Remaining is signal segment without friction.
5) setting radius parameter r is 1, calculates each test sample and all reference templates most using FastDTW algorithm The distance in good regular path.Fig. 4 illustrate first test sample and first reference template best regular path and its away from From.Lines in Fig. 4 from the lower left corner to the upper right corner represent the best regular path between two samples, the distance in best regular path Namely two shortest distances between sample, about 0.0005, therefore the similarity of two samples is very high；
6) according to Nearest neighbor rule, for each test sample, the label of the shortest reference template of selected distance is used as should The signal classification of test sample.The correct recognition rata for finally obtaining the test set is 95.4%.

Claims (1)

1. a kind of signal recognition method of the optical fiber perimeter defence system based on FastDTW algorithm, which is characterized in that including as follows Step:
1) the different types of output signal of optical fiber perimeter defence system is acquired in advance；
2) collected signal is divided into several signal segments according to classification；
3) each type of type signal segment is selected, it is normalized, as reference template, and to each mould Plate is marked；
4) freshly harvested signal to be identified is similarly divided and normalized, as test sample；
5) each test sample is calculated at a distance from the best regular path of all reference templates using FastDTW algorithm；
6) it according to Nearest neighbor rule, chooses with label of each test sample apart from shortest reference template as the test sample Signal classification.
CN201910336913.3A 2019-04-25 2019-04-25 The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm Pending CN110135283A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910336913.3A CN110135283A (en) 2019-04-25 2019-04-25 The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910336913.3A CN110135283A (en) 2019-04-25 2019-04-25 The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm

Publications (1)

Publication Number Publication Date
CN110135283A true CN110135283A (en) 2019-08-16

Family

ID=67570900

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910336913.3A Pending CN110135283A (en) 2019-04-25 2019-04-25 The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm

Country Status (1)

Country Link
CN (1) CN110135283A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110543907A (en) * 2019-08-29 2019-12-06 交控科技股份有限公司 fault classification method based on microcomputer monitoring power curve

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103235953A (en) * 2013-04-11 2013-08-07 北京交通大学 Pattern recognition method for fiber-optic distributed disturbance sensor
CN105825857A (en) * 2016-03-11 2016-08-03 无锡吾芯互联科技有限公司 Voiceprint-recognition-based method for assisting deaf patient in determining sound type
CN105973451A (en) * 2016-05-09 2016-09-28 深圳艾瑞斯通技术有限公司 Optical fiber vibration model determination method and device
CN106503642A (en) * 2016-10-18 2017-03-15 长园长通新材料股份有限公司 A kind of model of vibration method for building up for being applied to optical fiber sensing system
WO2017107494A1 (en) * 2015-12-25 2017-06-29 深圳市酷浪云计算有限公司 Method and device for recognizing badminton racket swinging motion
US20190018037A1 (en) * 2017-07-14 2019-01-17 Allstate Insurance Company Distributed data processing systems for processing remotely captured sensor data

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103235953A (en) * 2013-04-11 2013-08-07 北京交通大学 Pattern recognition method for fiber-optic distributed disturbance sensor
WO2017107494A1 (en) * 2015-12-25 2017-06-29 深圳市酷浪云计算有限公司 Method and device for recognizing badminton racket swinging motion
CN105825857A (en) * 2016-03-11 2016-08-03 无锡吾芯互联科技有限公司 Voiceprint-recognition-based method for assisting deaf patient in determining sound type
CN105973451A (en) * 2016-05-09 2016-09-28 深圳艾瑞斯通技术有限公司 Optical fiber vibration model determination method and device
CN106503642A (en) * 2016-10-18 2017-03-15 长园长通新材料股份有限公司 A kind of model of vibration method for building up for being applied to optical fiber sensing system
US20190018037A1 (en) * 2017-07-14 2019-01-17 Allstate Insurance Company Distributed data processing systems for processing remotely captured sensor data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
HAIWEI CAI ET AL.: "《Coherent detection based ϕ-OTDR and its application》" *
尹崇博 等: "《分布式光纤传感器触发模式识别的研究》" *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110543907A (en) * 2019-08-29 2019-12-06 交控科技股份有限公司 fault classification method based on microcomputer monitoring power curve

Similar Documents

Publication Publication Date Title
CN108968905A (en) Method, apparatus, system and the computer readable storage medium to give a test of one's eyesight
CN105022835A (en) Public safety recognition method and system for crowd sensing big data
CN111222743B (en) Method for judging vertical offset distance and threat level of optical fiber sensing event
CN105841793A (en) Optical fiber vibration source identification method, device and system
WO2018139887A1 (en) Method and device for adaptively configuring threshold for object detection by means of radar
CN109916434B (en) Optical fiber Fabry-Perot sensor cavity length multiple frequency related demodulation method
CN102879763A (en) System and method for quickly positioning noise source through image identification and sound intensity scanning
US4541113A (en) Apparatus and method of line pattern analysis
CN110135283A (en) The signal recognition method of optical fiber perimeter defence system based on FastDTW algorithm
CN101520509B (en) Method for comparatively measuring range of interframe of chaotic light time domain waveform
CN103903326A (en) Multi-spectral counterfeit bank note detection system and method
CN109613562A (en) Anemometry laser radar effectively detects recognition methods and the device on boundary
CN116863327B (en) Cross-domain small sample classification method based on cooperative antagonism of double-domain classifier
CN106650951B (en) Multi-channel measurement data automatic screening method based on time domain overall situation similarity
CN116489338A (en) Camera shooting delay test method and system
CN106767905B (en) Separate double detector type optical fibre gyro light source and electronic noise correlation calculations method
CN110424362A (en) A kind of optical fiber type temperature self-compensation static sounding sensor
US20150139539A1 (en) Apparatus and method for detecting forgery/falsification of homepage
CN105043544A (en) Image spectrum tracking method and system
Kamal et al. IOT Automation with Segmentation Techniques for Detection of Plant Seedlings in Agriculture
Rutigliano et al. Event-Detection Deep Neural Network for OTDR Trace Analysis
CN108985308A (en) The system and method for electric energy meter presentation quality is quickly analyzed based on data mining algorithm
CN114120509A (en) Regional access authorization device based on fusion identification technology
Ji et al. Influence of embedded microprocessor wireless communication and computer vision in Wushu competition referees’ decision support
Song et al. Extraction of double-peak Brillouin frequency shift based on ANN for nonuniform strain within BOTDR spatial resolution

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination