CN102853897A - Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) - Google Patents
Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) Download PDFInfo
- Publication number
- CN102853897A CN102853897A CN2012100740473A CN201210074047A CN102853897A CN 102853897 A CN102853897 A CN 102853897A CN 2012100740473 A CN2012100740473 A CN 2012100740473A CN 201210074047 A CN201210074047 A CN 201210074047A CN 102853897 A CN102853897 A CN 102853897A
- Authority
- CN
- China
- Prior art keywords
- module
- sampling
- signal
- fpga
- dsp
- 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.)
- Granted
Links
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention provides an optical fiber micro-vibration mode recognition system and an optical fiber micro-vibration mode recognition method on the basis of a digital signal processor (DSP) and a field programmable gate array (FPGA) and belongs to the technical field of optical fiber micro-vibration sensing. The optical fiber micro-vibration sensor mode recognition system on the basis of the DSP and the FPGA comprises a photovoltaic conversion module, an analog-digital conversion module, an FPGA module and a DSP module, the FPGA module comprises a sampling control module, a signal detection module and a first-in firs-out (FIFO) module, and the DSP module comprises a data exchange control module, a cache module, a characteristic extracting module, a classifier model training module and a classifier module testing module. The signal mode recognition system and the signal mode recognition method can achieve recognize types of output signals of a micro-vibration sensor, reduce false alarm rate of the optical fiber micro-vibration sensor, and achieve on-line real-time mode recognition of the signals of the optical fiber micro-vibration sensor.
Description
Technical field
The present invention relates to a kind of signal type recognition system and recognition methods thereof of Fiber optic micro-vibration sensor, belong to the little vibrating sensing technical field of optical fiber.
Background technology
Along with the development of optical fiber technology, Fiber optic micro-vibration sensor is applied in the systems such as circumference security protection, oil and natural gas pipeline and communication line monitoring more and more.Fiber optic micro-vibration sensor is to utilize optical fiber as a kind of distributed optical fiber sensing system of sensor information, and wherein optical fiber is both as sensor information, again as optical transmission medium.It can in sensor fibre is laid length, carry out long-range and real-time monitoring to the accident in certain accuracy range.Along with research and the application of various distributed optical fiber sensing systems, various disturbance locating methods be implemented in the position orientation problem that has solved to a certain extent intrusion alarm, but can't learn the main body of intrusion behavior at far end system.To the function of little vibration sensing system, the user is desirably on the basis, location of reporting to the police, and can determine to cause the active agent of reporting to the police, the signal that monitors is screened, reduce unnecessary warning, avoid the supervisor to be some unnecessary warning respondings, improve system effectiveness.
Existing research for the signal type of Fiber optic micro-vibration sensor is classified all is based on software analysis and the emulation of PC terminal, and the real-time of system is relatively poor.And in the application of Fiber optic micro-vibration sensor, such as circumference security protection, oil and gas pipeline monitoring and communication line monitoring etc., can in time carry out relevant treatment and export in real time the alerting signal of obtaining be the important component part of system performance, is the important indicator that can system effectively work.At PC end off-line the little vibrating sensing signal of optical fiber is carried out the requirement that type identification obviously can not satisfy on line real-time monitoring.
Summary of the invention
For problems of the prior art, the present invention proposes a kind of signal mode recognition device and recognition methods thereof of Fiber optic micro-vibration sensor, can realize the output signal type of vibrative sensor is identified, reduce the false alarm rate of Fiber optic micro-vibration sensor, realize the online real-time mode recognizing to the Fiber optic micro-vibration sensor signal.
The present invention proposes a kind of pattern recognition system of the Fiber optic micro-vibration sensor based on DSP and FPGA, comprises photoelectric conversion module, analog-to-digital conversion module, FPGA module and DSP module; Described FPGA module comprises controlling of sampling module, signal detection module and fifo module; Described DSP module data switching control module, cache module, characteristic extracting module, sorter model training module and sorter model test module.
Fiber optic micro-vibration sensor is to photoelectric conversion module output optical fibre vibrative sensor output signal, this signal process photoelectric conversion module is finished the transformation from the light intensity signal to the analog voltage signal, obtain analog voltage signal and export in the analog-to-digital conversion module, finish sampling to analog voltage signal by analog-to-digital conversion module, obtain to recover digital signal after the sampling of former Fiber optic micro-vibration sensor output signal principal character, realization will be converted into digital signal from simulating signal through the Fiber optic micro-vibration sensor output signal of opto-electronic conversion, analog-to-digital conversion module also is connected with controlling of sampling module in the FPGA module, the sampling process of control analog-to-digital conversion module; The rear digital signal of sampling transfers to signal detection module by analog-to-digital conversion module and carries out threshold decision, threshold decision adopts power spectrum method, the amplitude of the rear digital signal of sampling is carried out square, and square result is cumulative, when digital signal quantity after the cumulative sampling reaches the quantity corresponding with disturbance duration, accumulation result and threshold value are compared to judge whether former Fiber optic micro-vibration sensor output signal exists disturbance, when digital signal squared magnitude result surpasses threshold value after the cumulative sampling, judge that then there is disturbance in the Fiber optic micro-vibration sensor output signal, the rear digital signal of then sampling need input to fifo module and carry out the pattern classification processing, if less than threshold value, then there is not disturbance in former disconnected Fiber optic micro-vibration sensor output signal, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision.
After described fifo module is sampled rear digital signal and fills up, need transfer to the DSP module from the FPGA module by digital signal after the sampling that the disturbance of representing is arranged to exchanges data control module transmission look-at-me notice DSP module, after the exchanges data control module detects the look-at-me of fifo module, in the middle of digital signal directly read cache module from the fifo module in the FPGA module after the transmission mode of control DSP module startup EDMA will be sampled.
Described characteristic extracting module reads the digital signal after the sampling of storing in the cache module, digital signal after the controlling of sampling module samples is carried out wavelet decomposition, obtain the wavelet coefficient of this signal, and wavelet coefficient calculated, obtain the proper vector of this signal, send to the sorter model training module after the preservation, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal that represents disturbance, the sorter model training module utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model; After sorter support vector machine method model is determined, digital signal enters the sorter model test module from characteristic extracting module after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module, obtain the output of disturbance type identification.
The present invention also proposes the little vibration mode recognition methods of a kind of optical fiber based on DSP and FPGA, comprises following step:
Step 1: controlling of sampling process:
After FPGA module and DSP module power on, FPGA module and DSP module load respectively start-up routine, each subsystem is carried out initialization, then the FPGA module starts analog-to-digital conversion module, analog-to-digital conversion module is beginning under the control of controlling of sampling module sampling through the Fiber optic micro-vibration sensor output signal after the photoelectric conversion module conversion, the signal detection module of FPGA module adopts double-FIFO, an and FIFO in the middle of the interior double-FIFO of digital signal filling signal detection module after will sampling, after this FIFO is filled up by data, the real-time sampling passage is switched to another one FIFO proceed the signal filling, the notification signal detection module reads the FIFO that has filled up simultaneously, two FIFO rotate successively, realize the analog-to-digital conversion module real time data sampling, analog-to-digital conversion module also is connected with controlling of sampling module in the FPGA module, the sampling process of control analog-to-digital conversion module;
Step 2: micro-vibration signal testing process:
The rear digital signal of sampling transfers to signal detection module by analog-to-digital conversion module and carries out threshold decision, judge by the power spectrum that detects the rear digital signal of sampling whether former Fiber optic micro-vibration sensor output signal exists disturbance, when the squared magnitude accumulation result of digital signal after the sampling surpasses predefined threshold value, judge that then there is disturbance in the Fiber optic micro-vibration sensor output signal, the rear digital signal of then sampling need input to fifo module and carry out the pattern classification processing, if less than predefined threshold value, judge that then there is not disturbance in the Fiber optic micro-vibration sensor output signal, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision;
Step 3: DSP module and FPGA module data transmission procedure:
After fifo module is sampled rear digital signal and fills up, need transfer to the DSP module from the FPGA module by digital signal after the sampling that the disturbance of representing is arranged to exchanges data control module transmission look-at-me notice DSP module, after the exchanges data control module detects the fifo module look-at-me, in the middle of digital signal directly read cache module from the fifo module in the FPGA module after the transmission mode of control DSP module startup EDMA will be sampled;
Step 4: characteristic extraction procedure:
Characteristic extracting module reads the digital signal after the sampling of storing in the cache module, and the digital signal after the controlling of sampling module samples is carried out wavelet decomposition, obtains the wavelet coefficient of this signal, and wavelet coefficient is calculated, and obtains the proper vector of this signal;
Step 5: sorter model training process:
The proper vector of digital signal sends to the sorter model training module after the sampling after characteristic extracting module is preserved, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal that represents disturbance, the sorter model training module utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model, finally obtain representing the sorter support vector machine method model of the vibrative sensor output signal of various disturbance types;
Step 6: pattern classifier test process: after sorter support vector machine method model is determined, digital signal enters the sorter model test module from characteristic extracting module after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module, obtain the output of disturbance type identification.
The invention has the advantages that:
(1) the present invention proposes a kind of signal mode recognition device and recognition methods thereof of Fiber optic micro-vibration sensor, realization is to the pattern-recognition of Fiber optic micro-vibration sensor signal, increase the function of Fiber optic micro-vibration sensor, can understand the signal that triggers vibration at far-end and come Source Type;
(2) the present invention proposes a kind of signal mode recognition device and recognition methods thereof of Fiber optic micro-vibration sensor, realization is to the classification of Fiber optic micro-vibration sensor signal, to the signal of not wishing to report to the police, for example the warning that causes of natural cause wind and rain, hail is filtered, and reduces the false alarm rate of system;
(3) the present invention proposes a kind of signal mode recognition device and recognition methods thereof of Fiber optic micro-vibration sensor, the Virtex series of X C4VSX35 model FPGA superior performance of the TMS320C6747DSP of TI and Xilinx company, real-time, real-time implementation is to the classification of optical fiber micro-vibration signal fast, and real-time is superior;
(4) the present invention proposes a kind of signal mode recognition device and recognition methods thereof of Fiber optic micro-vibration sensor, realizes the little vibrational system miniaturization of optical fiber, and is close to the embedded system direction, breaks away from the restriction of PC end, realizes unmanned.
Description of drawings
Fig. 1: the functional block diagram of the signal mode recognition device of a kind of Fiber optic micro-vibration sensor that the present invention proposes;
Fig. 2: the signal mode recognition methods process flow diagram of a kind of Fiber optic micro-vibration sensor that the present invention proposes.
Among the figure: 1-Fiber optic micro-vibration sensor output signal; The 2-photoelectric conversion module; The 3-analog-to-digital conversion module;
4-controlling of sampling module; The 5-signal detection module; The 6-FIFO module;
7-exchanges data control module; The 8-cache module; The 9-characteristic extracting module;
10-sorter model training module; 11-sorter model test module; The output of 12-classification results;
The 13-FPGA module; The 14-DSP module.
Embodiment
The present invention is described in further detail below in conjunction with accompanying drawing.
The present invention proposes a kind of pattern recognition system of the Fiber optic micro-vibration sensor based on DSP and FPGA, comprises photoelectric conversion module 2, analog-to-digital conversion module 3, FPGA module 13 and DSP module 14.Described FPGA module 13 comprises controlling of sampling module 4, signal detection module 5 and fifo module 6 (First-In-First-Out module).Described DSP module 14 exchanges data control modules 7, cache module 8, characteristic extracting module 9, sorter model training module 10 and sorter model test module 11.
As shown in Figure 1, Fiber optic micro-vibration sensor is to photoelectric conversion module 2 output optical fibre vibrative sensor output signals 1, this signal process photoelectric conversion module 2 is finished the transformation from the light intensity signal to the analog voltage signal, obtain analog voltage signal and export in the analog-to-digital conversion module 3, finish sampling to analog voltage signal by analog-to-digital conversion module 3, obtain to recover digital signal after the sampling of original signal (being Fiber optic micro-vibration sensor output signal 1) principal character, realization will be converted into digital signal from simulating signal through the Fiber optic micro-vibration sensor output signal 1 of opto-electronic conversion, because only have digital signal to carry out the algorithm computing in the system that FPGA and DSP consist of.Described analog-to-digital conversion module 3 also is connected with controlling of sampling module 4 in the FPGA module 13, the sampling process of control analog-to-digital conversion module 3.The rear digital signal of sampling transfers to signal detection module 5 by analog-to-digital conversion module 3 and carries out threshold decision, threshold decision adopts power spectrum method, the amplitude of the rear digital signal of sampling is carried out square, cumulative square result, when digital signal quantity after the cumulative sampling reaches the quantity corresponding with disturbance duration, accumulation result and threshold value are compared to judge whether former Fiber optic micro-vibration sensor output signal 1 exists disturbance, (setting of threshold value can be adopted the method for statistics when digital signal squared magnitude result surpasses predefined threshold value after the cumulative sampling, the sampled signal squared magnitude corresponding disturbance adds up, for example about 50 disturbances, get its squared magnitude as a result the value of minimum as threshold value), judge that then there is disturbance in Fiber optic micro-vibration sensor output signal 1, the rear digital signal of then sampling need input to fifo module 6 and carry out the pattern classification processing, if less than predefined threshold value, then there is not disturbance in former Fiber optic micro-vibration sensor output signal 1, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision.
After described fifo module 6 is sampled rear digital signal and fills up, the sampling of the disturbance of represent is arranged by send look-at-mes notice DSP modules 14 to exchanges data control module 7 after the digital signal need transfer to DSP module 14 from FPGA module 13.After exchanges data control module 7 detects the look-at-me of fifo module 6, in the middle of digital signal directly read cache module 8 from the fifo module 6 in the FPGA module 13 after the transmission mode that control DSP module 14 starts EDMA will be sampled, in cache module 8, open up the onesize space of two and fifo module 6, place the corresponding data fifo (digital signal after the sampling) from FPGA module 13.The data of storage can directly read in the cache module 8, accelerate data access speed.
Described characteristic extracting module 9 reads the digital signal after the sampling of storing in the cache module 8, digital signal after 4 samplings of controlling of sampling module is carried out wavelet decomposition, obtain the wavelet coefficient of this signal, and wavelet coefficient calculated, obtain the proper vector of this signal, send to sorter model training module 10 after the preservation, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal 1 that represents disturbance, sorter model training module 10 utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model.Digital signal after the sampling is carried out the disturbance type identification and will be carried out two stages, set up at first according to the method described above sorter support vector machine method model, then rely on the method model to follow-up sampling after digital signal carry out the disturbance type identification.After sorter support vector machine method model is determined, digital signal enters sorter model test module 11 from characteristic extracting module 9 after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module 10, obtain the classification results output 12 of disturbance type identification.(function of sorter model training module 10 is to determine sorter model, and this need to accumulate the proper vector of some, for example the proper vector 20-50 of each disturbance type, determines sorter model based on these data; And rely on the sorter model of having determined in the sorter model training module 10 at sorter model test module 11, can carry out the disturbance type identification to digital signal after the follow-up sampling.)
Described FPGA module 13 has following functions: be connected control analog-to-digital conversion module 3 with analog-to-digital conversion module 33; Digital signal after the sampling is carried out threshold decision, and whether detection fiber vibrative sensor output signal 1 exists disturbance, whether needs to carry out pattern classification and processes; With the fifo module 6 that the digital signal after the sampling is sent to, notice DSP module 14 after FIFO is filled up by data.
Described DSP module 14 has following functions: be connected with FPGA module 13, the data communication by between 7 controls of exchanges data control module and the FPGA module 13 realizes that the digital signal after the sampling is transferred to cache module 8 by fifo module 6; Digital signal after the sampling is carried out wavelet decomposition, extract the signal wavelet coefficient; 3) find the solution information entropy according to wavelet coefficient, with this proper vector as the Fiber optic micro-vibration sensor signal that represents disturbance; 4) proper vector when the Fiber optic micro-vibration sensor signal that new representative disturbance is constantly arranged is gathered, proper vector according to the some (30-50) of the Fiber optic micro-vibration sensor of the different disturbances of the representative that collects, find the solution the optimal classification face based on support vector machine, determine the classifier algorithm model based on support vector machine; When the output signal of the new disturbance of Fiber optic micro-vibration sensor representative is gathered, can differentiate online the signal classification of the little vibrating sensing of optical fiber, as the people for trample, strong wind, hail etc.; If need to record the Fiber optic micro-vibration sensor signal that represents disturbance, can communicate by DSP module 14 built-in USB interface and PC ends.
What described analog-to-digital conversion module 3 adopted is that 12 potential differences are divided input AD9235, and simple and practical, sample frequency can effectively be restored the feature of original signal for being up to 20Msps.FPGA module 13 adopts the XC4VSX35 model of the Xilinx Virtex of company series, and the output signal to Fiber optic micro-vibration sensor that can be effectively real-time is carried out pre-service.The described DSP module 14 preferential TMS320C6747 that select TI company, its frequency outstanding performance is being arranged aspect the signal processing, and the resource on its chip is more up to 300M, comprise 2 EMIF interfaces, lcd controller and USB interface, be fit to realization and the expansion of system.Be integrated with USB interface, lcd controller in the DSP module 14, simple in structure, external corresponding power supply support just can realize the data high-speed transmission with the PC end, satisfies the transmission requirement of a large amount of real time datas of system, simplifies Interface design.
Described photoelectric conversion module 2 is preferably PINFET detecting module and the filtration module of SPF1200SF-D08 model, and this PINFET detecting module operation wavelength is 1000-1650nm, adopts positive and negative 5V power voltage supply.The detecting module of photoelectric conversion module 2 receives the signal that Fiber optic micro-vibration sensor spreads out of, and output signal outputs to analog to digital conversion circuit through filtration module (such as low-pass first order filter).
The sequential of described analog-to-digital conversion module 3 is by 13 controls of FPGA module, and the AD9235 that analog-to-digital conversion module 3 of the present invention adopts is that 12 potential differences are divided input analog-to-digital converter, can satisfy system requirements on the precision; FPGA module 13 is connected the circuit that connects with analog-to-digital conversion module have two groups of signals, the one, FPGA is to the control signal of analog-to-digital conversion module 3, contain enable signal CS and clock signal SCLK, the 2nd, data-signal SDATA, be specially the output digit signals of analog-to-digital conversion module 3, through analog-to-digital Fiber optic micro-vibration sensor output signal 1, when enable signal CS level is high, just enter FPGA module 13 from SDATA.
Be provided with fifo module 6 in the described FPGA module 13, link to each other by the EMIF interface with the cache module 8 of DSP module 14, FPGA module 13 is ready to sample after the rear numerical data, by interrupt notification DSP module 14, DSP module 14 adopts the EDMA mode to control data from the transmission of FPGA module 13 to DSP modules 14.There is abundant RAM resource FPGA module 13 inside of XC4VSX35 model, by the method for designing design fifo module 6 of IP kernel.The DSP module 14 of TMS320C6747 model is integrated two-way EMIF interface is respectively EMIFA and EMIFB, and wherein the EMIFA interface is the highest 16, and the EMIFB interface is the highest 32, only uses the EMIFA interface among the present invention.Cache module 8 by EMIFA Interface realization DSP module 14 links to each other with the fifo module 6 of FPGA module 13, the fifo module 6 of EMA_CS3 signal wire gating FPGA module 13 inside in the EMIFA interface in the DSP module 14, EMA_CLK, EMA_OE and EMA_WE, (EMA_CLK, EMA_OE and EMA_WE are the signal wire in the EMIFA interface) control is to the read-write sequence of fifo module 6; And link to each other with the GPIO pin of DSP module 14 from the INT look-at-me (fifo module 6 in the FPGA module 13 is sampled rear digital signal and fills up rear triggering INT look-at-me) of FPGA module 13, notice DSP module 14 starts the EDMA mode and carries out data transmission, EDMA module transmission mode makes DSP module 14 be absorbed in data-signal and processes not needing to realize data transmission in the DSP module 14 processing core intervention situations.
Among the present invention DSP module 14 adopt TMS320C6747 model DSP module 14 integrated the USB controller, comprise USB2.0OTG interface and USB 1.1OHCI (Host), support USB main equipment pattern and from equipment mode, support the control transmission, interrupt transmission, bulk transfer and etc. the time 4 types of transmission data transfer mode.The present invention adopts the OTG pattern of DSP module 14 internal USBs 2.0 modules, and TPS2065D provides power supply for VBUS by the current limliting panel switches.
The present invention also proposes the little vibration mode recognition methods of a kind of optical fiber based on DSP and FPGA, as shown in Figure 2, specifically comprise following step: the controlling of sampling process, the micro-vibration signal testing process, DSP and FPGA data transmission procedure, characteristic extraction procedure, sorter model training process, pattern classifier test process.
Step 1: controlling of sampling process:
The signal detection module 5 of FPGA module 13 adopts double-FIFO among the present invention, cooperates 3 pairs of Fiber optic micro-vibration sensor output signals of analog-to-digital conversion module 1 to carry out real-time sampling and in real time transfer.After FPGA module 13 and DSP module 14 powered on, FPGA module 13 and DSP module 14 loaded respectively start-up routine, and each subsystem is carried out initialization.Then FPGA module 13 starts analog-to-digital conversion module 3, analog-to-digital conversion module 3 is beginning under the control of controlling of sampling module 4 sampling through the Fiber optic micro-vibration sensor output signal 1 after photoelectric conversion module 2 conversions, an and FIFO in the middle of the digital signal filling signal detection module 5 interior double-FIFOs after will sampling, after this FIFO is filled up by data, the real-time sampling passage is switched to another one FIFO proceed the signal filling, notification signal detection module 5 reads the FIFO that has filled up simultaneously, two FIFO rotate successively, realize analog-to-digital conversion module 3 real time data samplings.Analog-to-digital conversion module 3 also is connected with controlling of sampling module 4 in the FPGA module 13, the sampling process of control analog-to-digital conversion module 3.
Step 2: micro-vibration signal testing process: the rear digital signal of sampling transfers to signal detection module 5 by analog-to-digital conversion module 3 and carries out threshold decision, judge by the power spectrum that detects the rear digital signal of sampling whether former Fiber optic micro-vibration sensor output signal 1 exists disturbance, when the squared magnitude accumulation result of digital signal after the sampling surpasses predefined threshold value, judge that then there is disturbance in Fiber optic micro-vibration sensor output signal 1, the rear digital signal of then sampling need input to fifo module 6 and carry out the pattern classification processing, if less than predefined threshold value, then there is not disturbance in former Fiber optic micro-vibration sensor output signal 1, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision.
Step 3: DSP module 14 and FPGA module 13 data distributing programs: after described fifo module 6 is sampled rear digital signal and fills up, the sampling of the disturbance of represent is arranged by send look-at-mes notice DSP modules 14 to exchanges data control module 7 after the digital signal need transfer to DSP module 14 from FPGA module 13.After exchanges data control module 7 detects fifo module 6 look-at-mes, in the middle of digital signal directly read cache module 8 from the fifo module 6 in the FPGA module 13 after the transmission mode that control DSP module 14 starts EDMA will be sampled, in cache module 8, open up the onesize space of two and fifo module 6, place the corresponding data fifo (digital signal after the sampling) from the FPGA module.The data of storage can directly read in the cache module 8, accelerate data access speed.
Step 4: characteristic extraction procedure: described characteristic extracting module 9 reads the digital signal after the sampling of storing in the cache module 8, digital signal after 4 samplings of controlling of sampling module is carried out wavelet decomposition, obtain the wavelet coefficient of this signal, and wavelet coefficient calculated, obtain the proper vector of this signal.
Step 5: sorter model training process: the proper vector of digital signal sends to sorter model training module 10 after the sampling after characteristic extracting module 9 is preserved, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal 1 that represents disturbance, sorter model training module 10 utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model.Finally obtain representing the sorter support vector machine method model of the vibrative sensor output signal of various disturbance types,
Step 6: pattern classifier test process: the digital signal after the sampling is carried out the disturbance type identification will carry out two stages, model sorter support vector machine method model, then just can rely on the method model to follow-up sampling after digital signal carry out the disturbance type identification.After sorter support vector machine method model is determined, digital signal enters sorter model test module 11 from characteristic extracting module 9 after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module 10, obtain the classification results output 12 of disturbance type identification.
Claims (7)
1. the pattern recognition system based on the Fiber optic micro-vibration sensor of DSP and FPGA is characterized in that: comprise photoelectric conversion module, analog-to-digital conversion module, FPGA module and DSP module; Described FPGA module comprises controlling of sampling module, signal detection module and fifo module; Described DSP module data switching control module, cache module, characteristic extracting module, sorter model training module and sorter model test module;
Fiber optic micro-vibration sensor is to photoelectric conversion module output optical fibre vibrative sensor output signal, this signal process photoelectric conversion module is finished the transformation from the light intensity signal to the analog voltage signal, obtain analog voltage signal and export in the analog-to-digital conversion module, finish sampling to analog voltage signal by analog-to-digital conversion module, obtain to recover digital signal after the sampling of former Fiber optic micro-vibration sensor output signal principal character, realization will be converted into digital signal from simulating signal through the Fiber optic micro-vibration sensor output signal of opto-electronic conversion, analog-to-digital conversion module also is connected with controlling of sampling module in the FPGA module, the sampling process of control analog-to-digital conversion module; The rear digital signal of sampling transfers to signal detection module by analog-to-digital conversion module and carries out threshold decision, threshold decision adopts power spectrum method, the amplitude of the rear digital signal of sampling is carried out square, and square result is cumulative, when digital signal quantity after the cumulative sampling reaches the quantity corresponding with disturbance duration, accumulation result and threshold value are compared to judge whether former Fiber optic micro-vibration sensor output signal exists disturbance, when digital signal squared magnitude result surpasses threshold value after the cumulative sampling, judge that then there is disturbance in former Fiber optic micro-vibration sensor output signal, the rear digital signal of then sampling need input to fifo module and carry out the pattern classification processing, if less than threshold value, judge that then there is not disturbance in the Fiber optic micro-vibration sensor output signal, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision;
After described fifo module is sampled rear digital signal and fills up, need transfer to the DSP module from the FPGA module by digital signal after the sampling that the disturbance of representing is arranged to exchanges data control module transmission look-at-me notice DSP module, after the exchanges data control module detects the look-at-me of fifo module, in the middle of digital signal directly read cache module from the fifo module in the FPGA module after the transmission mode of control DSP module startup EDMA will be sampled;
Described characteristic extracting module reads the digital signal after the sampling of storing in the cache module, digital signal after the controlling of sampling module samples is carried out wavelet decomposition, obtain the wavelet coefficient of this signal, and wavelet coefficient calculated, obtain the proper vector of this signal, send to the sorter model training module after the preservation, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal that represents disturbance, the sorter model training module utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model; After sorter support vector machine method model is determined, digital signal enters the sorter model test module from characteristic extracting module after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module, obtain the output of disturbance type identification.
2. the pattern recognition system of a kind of Fiber optic micro-vibration sensor based on DSP and FPGA according to claim 1, it is characterized in that: what described analog-to-digital conversion module adopted is that 12 potential differences are divided input AD9235, and sample frequency is 20Msps.
3. the pattern recognition system of a kind of Fiber optic micro-vibration sensor based on DSP and FPGA according to claim 1, it is characterized in that: described FPGA module adopts the XC4VSX35 model of Virtex series.
4. the pattern recognition system of a kind of Fiber optic micro-vibration sensor based on DSP and FPGA according to claim 1, it is characterized in that: described DSP module is selected TMS320C6747, and its frequency is 300M.
5. the pattern recognition system of a kind of Fiber optic micro-vibration sensor based on DSP and FPGA according to claim 1, it is characterized in that: described photoelectric conversion module is PINFET detecting module and the filtration module of SPF1200SF-D08 model, this PINFET detecting module operation wavelength is 1000-1650nm, adopts positive and negative 5V power voltage supply.
6. the pattern recognition system of a kind of Fiber optic micro-vibration sensor based on DSP and FPGA according to claim 1, it is characterized in that: the AD9235 that described analog-to-digital conversion module adopts is that 12 potential differences are divided input analog-to-digital converter.
7. little vibration mode recognition methods of the optical fiber based on DSP and FPGA is characterized in that: comprise following step:
Step 1: controlling of sampling process:
After FPGA module and DSP module power on, FPGA module and DSP module load respectively start-up routine, each subsystem is carried out initialization, then the FPGA module starts analog-to-digital conversion module, analog-to-digital conversion module is beginning under the control of controlling of sampling module sampling through the Fiber optic micro-vibration sensor output signal after the photoelectric conversion module conversion, the signal detection module of FPGA module adopts double-FIFO, an and FIFO in the middle of the interior double-FIFO of digital signal filling signal detection module after will sampling, after this FIFO is filled up by data, the real-time sampling passage is switched to another one FIFO proceed the signal filling, the notification signal detection module reads the FIFO that has filled up simultaneously, two FIFO rotate successively, realize the analog-to-digital conversion module real time data sampling, analog-to-digital conversion module also is connected with controlling of sampling module in the FPGA module, the sampling process of control analog-to-digital conversion module;
Step 2: micro-vibration signal testing process:
The rear digital signal of sampling transfers to signal detection module by analog-to-digital conversion module and carries out threshold decision, judge by the power spectrum that detects the rear digital signal of sampling whether former Fiber optic micro-vibration sensor output signal exists disturbance, when the squared magnitude accumulation result of digital signal after the sampling surpasses predefined threshold value, judge that then there is disturbance in former Fiber optic micro-vibration sensor output signal, the rear digital signal of then sampling need input to fifo module and carry out the pattern classification processing, if less than predefined threshold value, judge that then there is not disturbance in the Fiber optic micro-vibration sensor output signal, digital signal need not to process after the sampling, abandon and get final product, and wait for receive other samplings that arrive in turn after digital signals carry out threshold decision;
Step 3: DSP module and FPGA module data transmission procedure:
After fifo module is sampled rear digital signal and fills up, need transfer to the DSP module from the FPGA module by digital signal after the sampling that the disturbance of representing is arranged to exchanges data control module transmission look-at-me notice DSP module, after the exchanges data control module detects the fifo module look-at-me, in the middle of digital signal directly read cache module from the fifo module in the FPGA module after the transmission mode of control DSP module startup EDMA will be sampled;
Step 4: characteristic extraction procedure:
Characteristic extracting module reads the digital signal after the sampling of storing in the cache module, and the digital signal after the controlling of sampling module samples is carried out wavelet decomposition, obtains the wavelet coefficient of this signal, and wavelet coefficient is calculated, and obtains the proper vector of this signal;
Step 5: sorter model training process:
The proper vector of digital signal sends to the sorter model training module after the sampling after characteristic extracting module is preserved, find the solution information entropy, with the proper vector of information entropy as the Fiber optic micro-vibration sensor output signal that represents disturbance, the sorter model training module utilizes the proper vector of the different types of disturbing signal from Fiber optic micro-vibration sensor of the many groups of its resulting representative, method based on support vector machine, assess the optimal classification face between per two class signals, determine the topological structure of sorter, find the solution the optimal classification face of support vector machine, determine sorter support vector machine method model, finally obtain representing the sorter support vector machine method model of the vibrative sensor output signal of various disturbance types;
Step 6: pattern classifier test process: after sorter support vector machine method model is determined, digital signal enters the sorter model test module from characteristic extracting module after the follow-up sampling, under the sorter model control of having determined in the middle of the sorter model training module, obtain the output of disturbance type identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210074047.3A CN102853897B (en) | 2012-03-20 | 2012-03-20 | Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210074047.3A CN102853897B (en) | 2012-03-20 | 2012-03-20 | Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102853897A true CN102853897A (en) | 2013-01-02 |
| CN102853897B CN102853897B (en) | 2014-07-02 |
Family
ID=47400687
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210074047.3A Expired - Fee Related CN102853897B (en) | 2012-03-20 | 2012-03-20 | Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102853897B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103968933A (en) * | 2014-04-09 | 2014-08-06 | 西安电子科技大学 | Fuzzy domain characteristics based optical fiber vibration signal identifying method |
| CN105303777A (en) * | 2015-11-13 | 2016-02-03 | 中铁二院工程集团有限责任公司 | Tunnel lining structure spalling and collapsing real-time monitoring and alarming system |
| CN105928602A (en) * | 2016-04-15 | 2016-09-07 | 深圳艾瑞斯通技术有限公司 | Device and method for collecting and processing signal of optical fiber sensor |
| CN106992404A (en) * | 2016-12-15 | 2017-07-28 | 广东威创视讯科技股份有限公司 | A kind of intelligent power socket device |
| CN107510888A (en) * | 2017-08-17 | 2017-12-26 | 天津大学 | The local field potentials for removing DBS artefacts based on DSP+FPGA test and analyze platform in real time |
| CN109374110A (en) * | 2018-10-15 | 2019-02-22 | 北京航天控制仪器研究所 | A kind of self study device and method based on optical fiber vibration sensing system |
| CN110275596A (en) * | 2019-06-20 | 2019-09-24 | 深圳忆联信息系统有限公司 | Power-up initializing accelerated method, device and computer equipment based on solid state hard disk |
| CN110750484A (en) * | 2019-10-22 | 2020-02-04 | 西安因联信息科技有限公司 | Synchronous acquisition system and acquisition method for data of rotating speed and multiple vibration channels |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303788A (en) * | 2008-06-05 | 2008-11-12 | 北京航空航天大学 | Perimeter alarming method and system for implementing orientation of composite type complete optical fiber |
| JP2010085366A (en) * | 2008-10-02 | 2010-04-15 | Toshiba Corp | Apparatus for diagnosing abnormality in insulation of high-voltage electric instrument |
-
2012
- 2012-03-20 CN CN201210074047.3A patent/CN102853897B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303788A (en) * | 2008-06-05 | 2008-11-12 | 北京航空航天大学 | Perimeter alarming method and system for implementing orientation of composite type complete optical fiber |
| JP2010085366A (en) * | 2008-10-02 | 2010-04-15 | Toshiba Corp | Apparatus for diagnosing abnormality in insulation of high-voltage electric instrument |
Non-Patent Citations (3)
| Title |
|---|
| 刘吉等: "基于FPGA的高速数据处理系统设计", 《电子设计工程》 * |
| 罗光明等: "光纤振动传感器的信号检测电路设计", 《电子测量技术》 * |
| 蓝天等: "基于FPGA和DSP的全数字PGC解调系统设计", 《电子测量技术》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103968933A (en) * | 2014-04-09 | 2014-08-06 | 西安电子科技大学 | Fuzzy domain characteristics based optical fiber vibration signal identifying method |
| CN103968933B (en) * | 2014-04-09 | 2016-08-17 | 西安电子科技大学 | Fiber-optic vibration signal recognition method based on fuzzy field feature |
| CN105303777A (en) * | 2015-11-13 | 2016-02-03 | 中铁二院工程集团有限责任公司 | Tunnel lining structure spalling and collapsing real-time monitoring and alarming system |
| CN105928602A (en) * | 2016-04-15 | 2016-09-07 | 深圳艾瑞斯通技术有限公司 | Device and method for collecting and processing signal of optical fiber sensor |
| CN106992404A (en) * | 2016-12-15 | 2017-07-28 | 广东威创视讯科技股份有限公司 | A kind of intelligent power socket device |
| CN107510888A (en) * | 2017-08-17 | 2017-12-26 | 天津大学 | The local field potentials for removing DBS artefacts based on DSP+FPGA test and analyze platform in real time |
| CN109374110A (en) * | 2018-10-15 | 2019-02-22 | 北京航天控制仪器研究所 | A kind of self study device and method based on optical fiber vibration sensing system |
| CN110275596A (en) * | 2019-06-20 | 2019-09-24 | 深圳忆联信息系统有限公司 | Power-up initializing accelerated method, device and computer equipment based on solid state hard disk |
| CN110275596B (en) * | 2019-06-20 | 2020-12-29 | 深圳忆联信息系统有限公司 | Solid state disk-based power-on initialization acceleration method and device and computer equipment |
| CN110750484A (en) * | 2019-10-22 | 2020-02-04 | 西安因联信息科技有限公司 | Synchronous acquisition system and acquisition method for data of rotating speed and multiple vibration channels |
| CN110750484B (en) * | 2019-10-22 | 2022-11-25 | 西安因联信息科技有限公司 | Synchronous acquisition system and acquisition method for data of rotating speed and multiple vibration channels |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102853897B (en) | 2014-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102853897B (en) | Optical fiber micro-vibration mode recognition system and optical fiber micro-vibration mode recognition method on basis of digital signal processor (DSP) and field programmable gate array (FPGA) | |
| CN102001023A (en) | Tool wear detector | |
| CN102288346B (en) | Miniaturized digital large-scale sensor array impact monitoring system | |
| CN202373055U (en) | Embedded integrated lane controller | |
| CN103777609A (en) | Embedded intelligent monitoring and remote maintenance system of powder forming machine | |
| CN204340930U (en) | Slag-soil truck intelligent control system | |
| CN104199411A (en) | Industrial bus data gathering system based on Labview | |
| CN102506988B (en) | Monitoring system and method of on-line galloping and wind deflection of power transmission line | |
| CN202771495U (en) | Embedded intelligent traffic control toll terminal | |
| CN202033441U (en) | State detection device for valve chest of high-current rectifier | |
| CN206400756U (en) | Fast Ethernet data collecting card | |
| CN202915884U (en) | Winding deformation tester | |
| CN208333888U (en) | A kind of Portable train stationarity tester based on WIFI | |
| CN101639963B (en) | Implementation method of optical fiber vibration processor system | |
| CN204224015U (en) | Based on the embedded lift car person number detection device of ARM and double insurance elevator | |
| CN203432613U (en) | Wireless transmission power transmission line on-line monitoring system | |
| CN208012754U (en) | A kind of data collecting card suitable for optical fiber vibration sensing | |
| CN202836740U (en) | System for realizing remote monitoring on environmental noise | |
| CN204831538U (en) | Vibration control protection appearance | |
| CN206684798U (en) | A kind of Multifunctional attendance machine with PM2.5 monitoring functions | |
| CN203084719U (en) | SJA1000 interface IP core based on processor local bus (PLB) | |
| CN201393285Y (en) | Device for intelligent analysis and processing of video stream | |
| CN201465202U (en) | Multifunctional mechanical three-dimensional garage manipulator | |
| CN205353655U (en) | High integrated concurrent operation distribution monitoring and control system | |
| CN203773306U (en) | Multifunctional intelligent substation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20140702 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |