Summary of the invention
The technical problem that present invention mainly solves is to provide consistent optical fiber Recognition of Vibration Sources method, Apparatus and system, it is possible to real
Now improve accuracy rate and the speed identifying vibration source.
For solving above-mentioned technical problem, the technical scheme that the present invention uses is: provide a kind of optical fiber Recognition of Vibration Sources side
Method, is divided into J frame subsignal including: identification terminal by fiber-optic vibration signal;Extract the characteristic vector composition number of every frame subsignal
Group T [J]=T (0), T (j) ..., T (J-1) }, and obtain array R [I]={ R of the characteristic vector composition of preset signals model
(0), R (i) ..., R (I-1) }, wherein, the characteristic vector of described fiber-optic vibration signal and the Characteristic Vectors of described preset signals model
The extracting mode of amount is consistent;Determine distance g (R (0), T (0)) between described characteristic vector T (0) and described characteristic vector R (0)
And parameter M, wherein, the poor positive correlation between described M and described I and J;According to described distance g (R (0), T (0)), order meter
Calculate described array T [J] each characteristic vector T (j) respectively and between described array R [I] at least partly characteristic vector R (i) away from
From g (R (i), T (j)), until the distance g (R (I-being calculated between described characteristic vector T (J-1) and characteristic vector R (I-1)
1), T (J-1)), wherein, described g (R (i), T (j)) by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T
(j-1)) being calculated, described Partial Feature vector R (i) of each characteristic vector T (j) correspondence includes in described array R [I]
Characteristic vector R (max (j-M, 0)) to characteristic vector R (min (j+M, I-1));Calculate described distance g (R (I-1), T (J-1)) with
Described I and described J's and between ratio, using the similarity distance as described fiber-optic vibration signal Yu described preset signals model;
Impose a condition if described similarity distance meets, then the vibration source type of described fiber-optic vibration signal is defined as described preset signals mould
The vibration source type that type is corresponding.
Wherein, described g (R (i), T (j)) is by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-
1)) it is calculated and specifically includes: utilize formula 1 and formula 2 to be calculated described g (R (i), T (j));
Wherein, described characteristic vector T (j) is expressed as (y1..., yn), described characteristic vector R (i) is expressed as (x1...,
xn)。
Wherein, described distance g (R (0), T (0)) determined between described characteristic vector T (0) and described characteristic vector R (0)
Step include: utilize formula 3 to be calculated described distance g (R (0), T (0))
G (R (0), T (0))=2d (T (0), R (0)) (3).
Wherein, described according to described distance g (R (0), T (0)), order calculates described array T [J] each characteristic vector T
J the step of () distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i) including: root
According to described distance g (R (0), T (0)), order calculate described array T [J] each characteristic vector T (j) respectively with described characteristic vector
Distance g (R (0), T (j)) between R (0);Order calculate described array T [J] each characteristic vector T (j) respectively with described array
Distance g (R (i), T (j)) between R [I] at least partly characteristic vector R (i), wherein, as described j=0, described characteristic vector
Described Partial Feature vector R (i) corresponding for T (0) includes all characteristic vectors in described array R [I], when described j ≠ 0,
Described Partial Feature vector R (i) of described characteristic vector T (j) correspondence includes the characteristic vector R (max in described array R [I]
(j-M, 1)) to characteristic vector R (min (j+M, I-1)).
Wherein, described M=m+ | I-J |, m are a setting constant.
Wherein, the step of the characteristic vector of the every frame signal of described extraction includes: by described every frame subsignal respectively through line
Property predictive coding lpc analysis obtain correspondence cepstrum coefficient, using the cepstrum coefficient of described every frame subsignal as its characteristic vector.
Wherein, before the described step that fiber-optic vibration signal is divided into J frame subsignal, described method also includes: right
Described fiber-optic vibration signal carries out preemphasis process;After the described step that fiber-optic vibration signal is divided into J frame subsignal,
Described method also includes: every frame subsignal is carried out windowing process;Described every frame subsignal after described windowing process is entered
Row autocorrelation analysis.
For solving above-mentioned technical problem, another technical solution used in the present invention is: provide a kind of optical fiber Recognition of Vibration Sources
Device, including: divide module, for fiber-optic vibration signal being divided into J frame subsignal;Extraction module, is used for extracting every frame
Characteristic vector composition array T [J] of signal=T (0), T (j) ..., T (J-1) }, and obtain the Characteristic Vectors of preset signals model
Array R [I] of amount composition=R (0), R (i) ..., R (I-1) }, wherein, the characteristic vector of described fiber-optic vibration signal is with described
The extracting mode of the characteristic vector of preset signals model is consistent;First determines module, be used for determining described characteristic vector T (0) with
Distance g (R (0), T (0)) between described characteristic vector R (0) and parameter M, wherein, the difference between described M and described I and J
Positive correlation;First computing module, for according to described distance g (R (0), T (0)), order calculates described array T [J] each feature
Vector T (j) distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i), until calculating
Obtain distance g (R (I-1), T (J-1)) between described characteristic vector T (J-1) and characteristic vector R (I-1), wherein, described g (R
(i), T (j)) by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-1)) it is calculated, each feature
Described Partial Feature vector R (i) that vector T (j) is corresponding includes the characteristic vector R (max (j-M, 0)) in described array R [I] extremely
Characteristic vector R (min (j+M, I-1));Second computing module, is used for calculating described distance g (R (I-1), T (J-1)) and described I
And described J's and between ratio, using the similarity distance as described fiber-optic vibration signal Yu described preset signals model;Second is true
Cover half block, for when described similarity distance meets and imposes a condition, being defined as institute by the vibration source type of described fiber-optic vibration signal
State the vibration source type that preset signals model is corresponding.
Wherein, described g (R (i), T (j)) is by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-
1)) it is calculated and specifically includes: utilize formula 1 and formula 2 to be calculated described g (R (i), T (j));
Wherein, described characteristic vector T (j) is expressed as (y1..., yn), described characteristic vector R (i) is expressed as (x1...,
xn)。
For solving above-mentioned technical problem, another technical scheme that the present invention uses is: provide a kind of fiber identification system,
Including optical fiber, Fibre Optical Sensor and identification terminal;Described Fibre Optical Sensor is used for sending the first optical signal to described optical fiber one end,
And receive the second optical signal obtained by described first optical signal reflection from described optical fiber one end, and described second optical signal is entered
Row sampling obtains multiple Sampled optical signals, and the plurality of Sampled optical signals is converted to multiple sampling signal of telecommunication;Described identification
Terminal is used for determining when the first sampling signal of telecommunication is fiber-optic vibration signal, and described fiber-optic vibration signal is carried out Recognition of Vibration Sources, its
In, described identification terminal includes above-mentioned optical fiber Recognition of Vibration Sources device, so that described fiber-optic vibration signal is carried out Recognition of Vibration Sources.
Such scheme, processing terminal uses sliding window to sample by the fiber-optic signal obtaining Fibre Optical Sensor detection, and right
Sampled signal is identified, when recognizing sampled signal with preset signals Model Matching, by the environment shape of corresponding fiber position
Condition is defined as the default environmental aspect that this preset signals model is corresponding.This sliding window sample mode is i.e. every less than setting time span
Sampling time difference fiber-optic signal samples to obtain the sampled signal of this setting time span so that sampled signal comprises
Data abundanter, improve sampled signal and comprise the probability of enough valid data, when fiber-optic signal exist with preset letter
During the valid data of number Model Matching, above-mentioned sample mode the most greatly there may be sampled signal and comprises complete valid data,
Make by identifying that this sampled signal can accurately determine corresponding environmental aspect, therefore the identification that improve optic fibre environment situation is accurate
Degree.
Detailed description of the invention
Below describe in, in order to illustrate rather than in order to limit, it is proposed that such as particular system structure, interface, technology it
The detail of class, in order to thoroughly understand the application.But, it will be clear to one skilled in the art that and do not having these concrete
Other embodiment of details can also realize the application.In other situation, omit to well-known device, circuit with
And the detailed description of method, in order to avoid unnecessary details hinders the description of the present application.
Referring to Fig. 1, the flow chart of optical fiber Recognition of Vibration Sources method one embodiment of the present invention, the method includes:
Fiber-optic vibration signal is divided into J frame subsignal by S12: identification terminal.
Illustrating incorporated by reference to Fig. 2, Fig. 2 illustrates an optical fiber Recognition of Vibration Sources system, and this optical fiber Recognition of Vibration Sources system uses light
Impulse modulation system, changes, by the phase place of detection backscatter signals, the reflecting interference Strength Changes caused, it is possible to simultaneously
Detect multiple concurrent vibration source, thus realize early warning and vibration source is positioned.At this optical fiber Recognition of Vibration Sources system, Fibre Optical Sensor 21
It is connected with identification terminal 22.Optical fiber 23 is arranged in the environment that need to monitor such as underground, to monitor this environmental aspect.Fibre Optical Sensor
21 timings send the first optical signal from one end of optical fiber 23, and this first optical signal can be a pulse signal, as pulse width
For the laser of 10ns, this first optical signal the second optical signal that each position is formed through Rayleigh scattering in optical fiber 23, and
This second optical signal is reflected back one end of this optical fiber 23.Fibre Optical Sensor 21 obtains this second light letter from one end of this optical fiber 23
Number.Second optical signal is sampled by Fibre Optical Sensor 21, obtains multiple Sampled optical signals.Wherein, this sampling interval can gather
The optical signal that optical fiber is launched every setpoint distance, such as, first Sampled optical signals corresponds to apart from position, 1 meter, optical fiber one end anti-
The optical signal penetrated, second Sampled optical signals corresponds to the optical signal of the 2 meters of position reflections in distance optical fiber one end, by that analogy.
Due to the optical signal of backscattering and faint, and its noise is smaller, difficult during to optical signal prosessing
Degree is relatively big, precision is less, and therefore above-mentioned multiple Sampled optical signals are converted to the sampling signal of telecommunication of correspondence just by Fibre Optical Sensor 33
Process in signal.Here can be converted to analogue signal by general optical-electrical converter such as APD, then pass through analog digital conversion
Device converts analog signals into digital signal.
Multiple sampling signals of telecommunication after conversion are sent to identification terminal 22 by Fibre Optical Sensor 21.Certainly, implement at other
In example, this photoelectric conversion step can be performed by identification terminal 22, i.e. identification terminal 22 receives Fibre Optical Sensor 21 and believes the second light
Number multiple Sampled optical signals that sampling obtains, and convert thereof into multiple sampling signal of telecommunication.
Identification terminal 22 detects whether the sampling signal of telecommunication in the plurality of sampling signal of telecommunication is fiber-optic vibration signal, if
It is then to perform the step of this method embodiment.This fiber-optic vibration signal is expressed as occurring adopting of the fiber position reflection of vibration source
The fiber optic telecommunications number that sample optical signal is changed, which carry the vibration performance of this vibration source.
Characteristic vector composition array T [J] of S12: the identification terminal every frame subsignal of extraction=T (0), T (j) ..., T (J-
1) }, and obtain preset signals model characteristic vector composition array R [I]=R (0), R (i) ..., R (I-1).
Wherein, the extracting mode of the characteristic vector of the characteristic vector of described fiber-optic vibration signal and described preset signals model
Unanimously.
Such as, identification terminal storage has at least one preset signals model, this preset signals model correspondence each to include one
Multiple characteristic vector R (0) of the fiber-optic vibration signal of kind vibration source, R (i) ..., R (I-1), wherein, i is this preset signals model
The sequential label of signal frame, i=0 be this preset signals model play pip signal frame, i=I-1 is this preset signals model
Terminal subsignal frame, therefore I be the subsignal that this preset signals model comprises frame sum, R (i) is this preset signals mould
The characteristic vector of the subsignal of type the i-th frame.Identification terminal extract the 1st frame subsignal to J frame subsignal characteristic vector one by one
Sequentially correspond to T (0), T (j) ..., T (J-1), wherein, j is the sequential label of the signal frame of this fiber-optic vibration signal, and j=0 is
A pip signal frame of this fiber-optic vibration signal, j=J-1 is the terminal subsignal frame of this fiber-optic vibration signal, and therefore J is this light
The frame sum of the subsignal that fine vibration signal is comprised, T (i) is the Characteristic Vectors of the subsignal of this fiber-optic vibration signal jth frame
Amount.Above-mentioned I and J is all higher than 1, and both can be equal or unequal, in this no limit.
It should be noted that this identification terminal extracts mode and the spy in preset signals model of the characteristic vector of subsignal
The extracting mode levying vector is consistent, to ensure that both accurate following contrasts.That is, preset signals model and fiber-optic vibration
Signal uses the characteristic vector of same type.
Wherein, its extracting mode can be multiple, for example, linear predictive coding (linear predictive
Coding, LPC) parameter that can represent this subsignal feature that obtains, such as LPC coefficient or cepstrum coefficient.In another embodiment
In, the step of the characteristic vector of the every frame signal of described extraction includes: it is right to be obtained respectively through lpc analysis by described every frame subsignal
The cepstrum coefficient answered, using the cepstrum coefficient of described every frame subsignal as its characteristic vector.
S13: identification terminal determines distance g (R (0), T between described characteristic vector T (0) and described characteristic vector R (0)
) and parameter M (0).
Wherein, the poor positive correlation between described M and described I and J.Such as, described M=m+ | I-J |, m are a setting constant.
This m can require to be correspondingly arranged according to computing etc., to reduce the operand of following step S14, optimizes the fortune of following step S14
Line speed.Such as being set to the least as m, the operand of following step S14 is the fewest.In one specifically application, this m may be configured as I
Or 1 to a thirtieth/10th of J, and less than 10.
In the present embodiment, identification terminal utilize formula 11 be calculated characteristic vector T (0) and described characteristic vector R (0) it
Between distance g (R (0), T (0)).
G (R (0), T (0))=2d (T (0), R (0)) (11)
Wherein, the definition of this d refers to formula 13 and the associated description thereof of step S14.
S14: identification terminal calculates described array T [J] each characteristic vector according to described distance g (R (0), T (0)), order
The T (j) distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i), until being calculated
Distance g (R (I-1), T (J-1)) between described characteristic vector T (J-1) and characteristic vector R (I-1).
Wherein, described g (R (i), T (j)) is by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-
1)) it is calculated.Such as, identification terminal utilizes formula 12 and formula 13 to be calculated described g (R (i), T (j));
Wherein, described characteristic vector T (j) is expressed as (y1..., yn), described characteristic vector R (i) is expressed as (x1..., xn)。
Certainly, in other embodiments, this distance function d may be used without Euclidean distance, for
Wherein, described Partial Feature vector R (i) that each characteristic vector T (j) is corresponding includes the spy in described array R [I]
Levy vector R (max (j-M, 0)) to characteristic vector R (min (j+M, I-1)).
Said sequence calculates and is represented by: according to the order of elements of array T [J], calculates each characteristic vector T (j) with same
The distance of one characteristic vector R (i), and according to the order of elements of array R [I], calculate its characteristic vector R (i) and same Characteristic Vectors
The distance of amount T (j).Such as above-mentioned formula 12, each characteristic vector T (j) is special before need to relying on it with the distance of characteristic vector R (i)
Levy the distance between vector, therefore need to calculate according to array order.
Referring to Fig. 3, in another embodiment, this S14 specifically includes following sub-step:
S141: identification terminal calculates described array T [J] each Characteristic Vectors according to described distance g (R (0), T (0)), order
Amount T (j) distance g (R (0), T (j)) respectively and between described characteristic vector R (0).
S142: identification terminal order calculates described array T [J] each characteristic vector T (j) respectively with described array R [I] extremely
Distance g (R (i), T (j)) between small part characteristic vector R (i).
Wherein, as described j=0, described Partial Feature vector R (i) of described characteristic vector T (0) correspondence includes described
All characteristic vectors in array R [I], when described j ≠ 0, the described Partial Feature vector R that described characteristic vector T (j) is corresponding
I () includes the characteristic vector R (max (j-M, 1)) to characteristic vector R (min (j+M, I-1)) in described array R [I].
For improve computing degree of accuracy, identification terminal first calculate described array T [J] each characteristic vector T (j) respectively with institute
State distance g (R (0), T (j)) between characteristic vector R (0), and characteristic vector T (0) respectively with described array R [I] each spy
Levy distance g (R (i), T (0)) between vector R (i).Then, identification terminal order again calculates in described array T [J] except T (0)
Outer each characteristic vector T (j) distance g (R (i), T respectively and between described array R [I] at least partly characteristic vector R (i)
(j)), described Partial Feature vector R (i) of described characteristic vector T (j) correspondence includes the characteristic vector R in described array R [I]
(max (j-M, 1)) to characteristic vector R (min (j+M, I-1)).Though this embodiment has been slightly increased characteristic vector R (0), Characteristic Vectors
The distance operation amount of amount T (0), but computing accuracy can be improved further.
Below above-mentioned calculation is explained, in the present invention, in order to compare fiber-optic vibration signal and preset letter
Similarity between number model, uses the distance calculated between them, and the least then similarity of distance is the highest.Shake to calculate optical fiber
Dynamic distance between signal and preset signals model, need to from fiber-optic vibration signal and preset signals model each corresponding subsignal
Distance between frame is counted.
Such as, the calculating of the similarity distance between fiber-optic vibration signal and preset signals model can use dynamic programming (DP)
Method.Concrete the most such as set up two-dimensional direct angle coordinate system, if each frame number j=0 of fiber-optic vibration signal~(J-1) at transverse axis
On mark, each frame number i=0~(I-1) of preset signals model are marked, respectively to represent frame number on transverse and longitudinal axle on the longitudinal axis
Each coordinate makes the straight line being perpendicular to this coordinate place axle as initial point, thus can form a network, in this network
(i j) represents the joint of a certain frame of correspondence of fiber-optic vibration signal and preset signals model in each cross point.DP algorithm can
Be attributed to find one by the path of some lattice points in this network, the lattice point that path is passed through is fiber-optic vibration signal and pre-
If signal model carries out the frame number calculated.Path is not elective, and the speed of first any signal is likely to
Change, but the precedence of its each several part it is impossible to change that, therefore selected path must be to go out from the lower left corner (i, j=0)
Send out, terminate at the upper right corner (i=I-1, j=J-1).
Because fiber-optic vibration signal is different with preset signals model length, so the matching relationship of its correspondence has a variety of,
Therefore need to find out distance that coupling path the shortest, when from above-mentioned network one grid (i-1, j-1), (i-1, j) or
(i, j-1) move to next grid (i, j), if if transversely or longitudinally mobile, its distance for d (i, j), if
What diagonal came sideling is then 2d (i, j), so, fiber-optic vibration signal and two characteristic vector times of preset signals model
The most above-mentioned formula of distance 13.This g (R (i), T (j)) represents that fiber-optic vibration signal and preset signals model are all vowed from initiation feature
Amount is gradually matched to characteristic vector R (i) and characteristic vector T (j).When being matched to characteristic vector R (I-1) and characteristic vector T (J-1)
Distance g (R (I-1), T (J-1)), i.e. matching this step is the distance between fiber-optic vibration signal and preset signals model.
S15: identification terminal calculate described distance g (R (I-1), T (J-1)) and described I and described J's and between ratio, with
Similarity distance as described fiber-optic vibration signal Yu described preset signals model.
Such as, identification terminal obtains being matched to distance g (R (I-1), the T of characteristic vector R (I-1) and characteristic vector T (J-1)
(J-1), after), the similarity distance s of described fiber-optic vibration signal and described preset signals model it is calculated according to formula 14;
S16: impose a condition if described similarity distance meets, then identification terminal is by the vibration source type of described fiber-optic vibration signal
It is defined as the vibration source type that described preset signals model is corresponding.
This impose a condition as less than set similarity distance, or be the minimum in all preset signals models similar away from
From.Such as, identification terminal stores multiple preset signals models, and identification terminal is performed a plurality of times above-mentioned steps S13-S15, obtains each
Preset signals model and the similarity distance of this fiber-optic vibration model, by identification terminal by the vibration source classification of type of fiber-optic vibration model
For the vibration source type that the preset signals model that similarity distance is minimum is corresponding.Certainly, for the different demands of concrete application, this setting
Condition can be not especially limited at this with other conditions.
In the present embodiment, identification terminal is believed with presetting by calculating the characteristic vector of every frame subsignal of fiber-optic vibration signal
The distance of the characteristic vector of number model, to obtain the similarity between this fiber-optic vibration signal speech preset signals model, and then
The vibration source type of fiber-optic vibration signal is determined, it is achieved that the vibration source of fiber-optic vibration signal is classified by similarity, and this classification side
Formula can carry out Accurate classification to vibration source, improves the accuracy rate of Recognition of Vibration Sources, and identification terminal only calculates according to setting rule
Distance between array T [J] each characteristic vector T (j) and array R [I] Partial Feature vector R (i), decreases operand, carries
High recognition speed and efficiency, save the process resource of identification terminal.
Refer to the partial process view that Fig. 4, Fig. 4 are optical fiber Recognition of Vibration Sources method a further embodiments of the present invention, the method
Including above-mentioned S12-S16 and further comprising the steps of before step S12:
S41: identification terminal carries out preemphasis process to described fiber-optic vibration signal.
Such as, the preemphasis network of employing is a fixing single order digital display circuit, and this signal equation is:
S (n)=s (n)-0.94s (n-1);
Wherein, s (n) is fiber-optic vibration signal before treatment, and S (n) is fiber-optic vibration signal before treatment.
Fiber-optic vibration signal is divided into J frame subsignal by S42: identification terminal.
S43: identification terminal carries out windowing process to every frame subsignal.
Such as, Hamming (Hamming) window is used to eliminate the sharp change of the signal edge caused by framing, in one embodiment,
The definition of this Hamming window is:
S44: identification terminal carries out autocorrelation analysis to the described every frame subsignal after described windowing process.
Such as, identification terminal carries out autocorrelation calculation to the described every frame subsignal after described windowing process and is:Wherein, the subsignal after S represents windowing process.
Described every frame subsignal is obtained the LPC coefficient on p rank by S45: identification terminal respectively through lpc analysis, by described often
The LPC coefficient on the p rank of frame subsignal is converted to the cepstrum coefficient on the q rank of correspondence.
The cepstrum coefficient on the q rank of this every frame subsignal is the characteristic vector of this subsignal, and above-mentioned p, q are all not equal to 0.On
State and obtain LPC coefficient i.e. cepstrum coefficient by lpc analysis and see existing correlation analysis, be not specifically described at this.In this reality
Executing in example, the characteristic vector of above-mentioned preset signals model is also the cepstrum coefficient of same type, and this preset signals model uses same
The forms of sample carry out windowing process.
Referring to the structural representation that Fig. 5, Fig. 5 are Recognition of Vibration Sources device one embodiments of the present invention, this device includes:
Framing module 51, for being divided into J frame subsignal by fiber-optic vibration signal.
Extraction module 52, for extract every frame subsignal characteristic vector composition array T [J]=T (0), T (j) ..., T
(J-1) }, and obtain preset signals model characteristic vector composition array R [I]=R (0), R (i) ..., R (I-1), its
In, the characteristic vector of described fiber-optic vibration signal is consistent with the extracting mode of the characteristic vector of described preset signals model.
First determines module 53, is used for determining distance g between described characteristic vector T (0) and described characteristic vector R (0)
(R (0), T (0)) and parameter M, wherein, the poor positive correlation between described M and described I and J.
First computing module 54, for according to described distance g (R (0), T (0)), it is each that order calculates described array T [J]
Characteristic vector T (j) distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i), until
Distance g (R (I-1), T (J-1)) being calculated between described characteristic vector T (J-1) and characteristic vector R (I-1), wherein, institute
State g (R (i), T (j)) by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-1)) it is calculated, each
Described Partial Feature vector R (i) that characteristic vector T (j) is corresponding include characteristic vector R in described array R [I] (max (j-M,
0)) to characteristic vector R (min (j+M, I-1)).
Second computing module 55, be used for calculating described distance g (R (I-1), T (J-1)) and described I and described J's and between
Ratio, using the similarity distance as described fiber-optic vibration signal Yu described preset signals model.
Second determines module 56, for when described similarity distance meets and imposes a condition, by described fiber-optic vibration signal
Vibration source type is defined as the vibration source type that described preset signals model is corresponding.
Optionally, described g (R (i), T (j)) is by g (R (i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T
(j-1)) it is calculated and specifically includes: the first computing module 54 utilizes above-mentioned formula 12 and formula 13 to be calculated described g (R
(i),T(j))。
Alternatively, described distance g (R (0), the T determined between described characteristic vector T (0) and described characteristic vector R (0)
(0) step) includes: first determines that module 53 utilizes above-mentioned formula 11 to be calculated described distance g (R (0), T (0)).
Alternatively, the first computing module 54 performs described according to described distance g (R (0), T (0)), and order calculates described number
Group T [J] each characteristic vector T (j) distance g respectively and between described array R [I] at least partly characteristic vector R (i) (R (i),
T (j)) step include: according to described distance g (R (0), T (0)), order calculates described array T [J] each characteristic vector T (j)
Distance g (R (0), T (j)) respectively and between described characteristic vector R (0);Order calculates described array T [J] each characteristic vector
The T (j) distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i), wherein, as described j
When=0, described Partial Feature vector R (i) of described characteristic vector T (0) correspondence includes all features in described array R [I]
Vector, when described j ≠ 0, described Partial Feature vector R (i) of described characteristic vector T (j) correspondence includes described array R [I]
In characteristic vector R (max (j-M, 1)) to characteristic vector R (min (j+M, I-1)).
Alternatively, described M=m+ | I-J |, m are a setting constant.
Alternatively, the step of the characteristic vector that extraction module 52 performs the every frame signal of described extraction includes: by described every frame
Subsignal obtains the cepstrum coefficient of correspondence respectively through linear predictive coding lpc analysis, by the cepstrum system of described every frame subsignal
Number is as its characteristic vector.
Alternatively, this device also includes:
Pre-emphasis module, the optical fiber for described fiber-optic vibration signal is carried out preemphasis process, after preemphasis is processed
Vibration signal sends to framing module 51.
Windowing module, carries out windowing process for what framing module 51 exported to every frame subsignal;
Auto-correlation module, for carrying out autocorrelation analysis to the described every frame subsignal after described windowing process.
Wherein, the above-mentioned module of this identification device is respectively used to perform the corresponding steps in said method embodiment, specifically
Execution process as above embodiment of the method illustrates, therefore not to repeat here.
Being the structural representation of Recognition of Vibration Sources another embodiment of device of the present invention refering to Fig. 6, Fig. 6, this device 60 includes
Processor 61, memorizer 62, receptor 63 and bus 64.Wherein, processor 61, memorizer 62, receptor 63 may each be one
Individual or multiple, in Fig. 6 only as a example by one.
Receptor 63 is for receiving the information that external equipment sends.Such as, receive Fibre Optical Sensor detection obtain by many
The individual sampling signal of telecommunication.
Memorizer 62 is used for storing computer program, and provides described computer program to processor 61, and can store place
The data that reason device 61 processes, multiple sampling signals of telecommunication that such as receptor 63 receives.Wherein, memorizer 62 can include read-only
At least one in memorizer, random access memory and nonvolatile RAM (NVRAM).
In embodiments of the present invention, processor 61, by performing the computer program of memorizer 62 storage, is used for:
Fiber-optic vibration signal is divided into J frame subsignal;Such as, to be that receptor 63 receives many for this fiber-optic vibration signal
One of them sampling signal of telecommunication in the individual sampling signal of telecommunication.Processor can carry out vibration detection to the plurality of sampling signal of telecommunication,
Determine when one of them sampling signal of telecommunication is fiber-optic vibration signal, this fiber-optic vibration signal is carried out above-mentioned framing.
Extract characteristic vector composition array T [J] of every frame subsignal=T (0), T (j) ..., T (J-1), and obtain pre-
If array R [I] of the characteristic vector composition of signal model=R (0), R (i) ..., R (I-1) }, wherein, described fiber-optic vibration is believed
Number characteristic vector consistent with the extracting mode of the characteristic vector of described preset signals model;
Determine distance g (R (0), T (0)) between described characteristic vector T (0) and described characteristic vector R (0) and parameter
M, wherein, the poor positive correlation between described M and described I and J;
According to described distance g (R (0), T (0)), order calculate described array T [J] each characteristic vector T (j) respectively with institute
State distance g (R (i), T (j)) between array R [I] at least partly characteristic vector R (i), until being calculated described characteristic vector
Distance g (R (I-1), T (J-1)) between T (J-1) and characteristic vector R (I-1), wherein, described g (R (i), T (j)) is by g (R
(i-1), T (j)), g (R (i-1), T (j-1)) or g (R (i), T (j-1)) be calculated, each characteristic vector T (j) is corresponding
Described Partial Feature vector R (i) includes the characteristic vector R (max (j-M, 0)) to characteristic vector R (min in described array R [I]
(j+M,I-1));
Calculate described distance g (R (I-1), T (J-1)) and described I and described J's and between ratio, using as described optical fiber
Vibration signal and the similarity distance of described preset signals model;
Impose a condition if described similarity distance meets, then the vibration source type of described fiber-optic vibration signal is defined as described pre-
If the vibration source type that signal model is corresponding.
Alternatively, processor 61 perform described g (R (i), T (j)) by g (R (i-1), T (j)), g (R (i-1), T (j-1)),
Or g (R (i), T (j-1)) is calculated and specifically includes: utilize above-mentioned formula 12 and formula 13 to be calculated described g (R (i), T
(j))。
Alternatively, processor 61 perform described determine between described characteristic vector T (0) and described characteristic vector R (0) away from
Step from g (R (0), T (0)) includes: utilize above-mentioned formula 11 to be calculated described distance g (R (0), T (0)).
Alternatively, processor 61 performs described according to described distance g (R (0), T (0)), and order calculates described array T [J]
Each characteristic vector T (j) distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i)
Step include: according to described distance g (R (0), T (0)), order calculates described array T [J] each characteristic vector T (j) respectively
And distance g (R (0), T (j)) between described characteristic vector R (0);Order calculates described array T [J] each characteristic vector T (j)
Distance g (R (i), T (j)) respectively and between described array R [I] at least partly characteristic vector R (i), wherein, as described j=0
Time, described Partial Feature vector R (i) of described characteristic vector T (0) correspondence includes all Characteristic Vectors in described array R [I]
Amount, when described j ≠ 0, described Partial Feature vector R (i) of described characteristic vector T (j) correspondence includes in described array R [I]
Characteristic vector R (max (j-M, 1)) to characteristic vector R (min (j+M, I-1)).
Alternatively, described M=m+ | I-J |, m are a setting constant.
Alternatively, the step of the characteristic vector that processor 61 performs the every frame signal of described extraction includes: by described every frame
Signal obtains the cepstrum coefficient of correspondence respectively through linear predictive coding lpc analysis, by the cepstrum coefficient of described every frame subsignal
As its characteristic vector.
Alternatively, processor 61, before performing the described step that fiber-optic vibration signal is divided into J frame subsignal, is also used
In described fiber-optic vibration signal is carried out preemphasis process;
After performing the described step that fiber-optic vibration signal is divided into J frame subsignal, it is additionally operable to: to every frame subsignal
Carry out windowing process;Described every frame subsignal after described windowing process is carried out autocorrelation analysis.
Above-mentioned processor 61 can also be referred to as CPU (Central Processing Unit, CPU).Specifically
Application in, each assembly of terminal is coupled by bus 64, and wherein bus 64 is in addition to including data/address bus, also may be used
To include power bus, to control bus and status signal bus in addition etc..But for the sake of understanding explanation, in the drawings by various buses
All it is designated as bus 64.The method that the invention described above embodiment discloses can also be applied in processor 61, or by processor
61 realize.
In such scheme, identification terminal is believed with presetting by calculating the characteristic vector of every frame subsignal of fiber-optic vibration signal
The distance of the characteristic vector of number model, to obtain the similarity between this fiber-optic vibration signal speech preset signals model, and then
The vibration source type of fiber-optic vibration signal is determined, it is achieved that the vibration source of fiber-optic vibration signal is classified by similarity, and this classification side
Formula can carry out Accurate classification to vibration source, improves the accuracy rate of Recognition of Vibration Sources, and identification terminal only calculates according to setting rule
Distance between array T [J] each characteristic vector T (j) and array R [I] Partial Feature vector R (i), decreases operand, carries
High recognition speed and efficiency, save the process resource of identification terminal.
The foregoing is only embodiments of the present invention, not thereby limit the scope of the claims of the present invention, every utilization is originally
Equivalent structure or equivalence flow process that description of the invention and accompanying drawing content are made convert, or are directly or indirectly used in what other were correlated with
Technical field, is the most in like manner included in the scope of patent protection of the present invention.