CN102692625A - Joint modeling method for features of underwater target echo and reverberation in Rn space - Google Patents
Joint modeling method for features of underwater target echo and reverberation in Rn space Download PDFInfo
- Publication number
- CN102692625A CN102692625A CN2012101500725A CN201210150072A CN102692625A CN 102692625 A CN102692625 A CN 102692625A CN 2012101500725 A CN2012101500725 A CN 2012101500725A CN 201210150072 A CN201210150072 A CN 201210150072A CN 102692625 A CN102692625 A CN 102692625A
- Authority
- CN
- China
- Prior art keywords
- space
- signal
- reverberation
- target echo
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention provides a joint modeling method for features of underwater target echo and reverberation in an Rn space. The joint modeling method includes steps of firstly, building a training sample bank containing target echo and reverberation, and mapping training samples to multiple signal feature spaces by different signal processing methods according differences of physical properties of two types of signals; secondly, reducing dimensionality of the signal feature spaces according to distribution information of the training samples in the signal feature spaces, fusing the different signal feature spaces to build a fusion signal feature space; thirdly, building a joint distribution model of the target echo and reverberation of the training samples in the fusion signal feature space; fourthly, mapping unknown signal samples into the fusion signal feature space by the signal processing methods in the first step and the second step, and comparing to the joint distribution model built in the third step to judge types of the unknown signal samples. The feature space built by the joint modeling method has the advantage of enabling separability of target echo and reverberation to be more stable and universal.
Description
Technical field
The present invention relates to the underwater acoustic technology application, specifically a kind of R
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation.
Background technology
Sink to the bottom under water or bury in the underwater operation tasks such as target detection is engaged in archaeological studies under water, raising of a wreck and have important effect.Based on detection range, water-bed target acquisition can be divided into proximity detection and two kinds of situation of long-range detection.During proximity detection, can adopt equipment such as imaging sonar, obtain target and near acoustic picture thereof, intuitively target discerned and judged.And during long-range detection, target does not possess image-forming condition, can't obtain the shape information of target, can only whether exist target and judges according to being modulated at target vibration information in the echoed signal.
When carrying out remote water-bed target detection at present, the way that adopts usually is to use wideband pulse of active sonar emission, and adopts sonar array to receive echoed signal.According to different on Physical Mechanism of target echo and reverberation; Confirm the two difference on signal properties; And select corresponding mathematic(al) manipulation method to extract the signal characteristic of echoed signal, adopt mode identification method that the signal characteristic of echoed signal is judged identification.In this signal processing flow, present research and achievement mainly concentrate on to be analyzed the character of target echo and reverberation, and studies corresponding signal characteristic extracting methods.In that this side up; Domestic scholars has proposed to be applicable to the target echo bright spots model of practical applications; The clear and definite citation form of how much echoes and elasticity echo in the target echo, and adopted multiple signal processing method to extract signal characteristic in view of the above based on time domain, frequency domain and time-frequency domain.But the method for having studied at present all is when extracting signal characteristic, reverberation to be suppressed as a kind of undesired signal, does not see the pertinent literature report to the signal properties of reverberation and the research of characteristic thereof.
Summary of the invention
The objective of the invention is to propose a kind of optimum separability that under the Euclidean distance meaning, has, can improve the detection performance of active sonar, realize the R that telesecurity is surveyed
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation.
The objective of the invention is to realize like this:
A kind of R
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation may further comprise the steps:
Step 1: with reverberation as one type of signal source with stabilization signal character; Foundation comprises the training sample database of target echo and reverberation; Based on two kinds of signal differences of physical properties, adopt the unlike signal processing method that training sample is mapped in a plurality of signal characteristics space;
Step 2:,, and, set up and merge the signal characteristic space to the fusion of various signals feature space to signal characteristic space dimensionality reduction according to the distributed intelligence of training sample in the signal characteristic space;
Step 3: set up target echo and the joint distribution model of reverberation in merging the signal characteristic space in the training sample;
Step 4: for the sample of signal of the unknown, be mapped in the fusion feature space, and compare, judge the kind of unknown signaling sample with joint distribution model that step 3 is set up through step 1 and the described signal processing method of step 2.
Described employing unlike signal disposal route is meant and adopts Fourier Transform of Fractional Order that training sample is mapped to energy accumulating property feature space and adopts the Hilbert-Huang conversion that training sample is mapped to many components feature space.
Described signal characteristic space dimensionality reduction, employing be under the Euclidean distance meaning, carry out based on the linear discriminant analysis of Fisher criterion function.
Merge in described signal characteristic space, employing be that canonical correlation analysis combines the method for series connection Feature Fusion to carry out.
The modeling of described feature space joint distribution, employing be that the discriminant function method is set up math equation to target echo and the classifying face of reverberation in the fusion feature space.
The principal feature of method of the present invention:
The present invention is regarded as one type of signal source with stabilization signal character with reverberation, according to target echo and the difference of reverberation on signal properties, in conjunction with feature space compression and fusion method, has set up a fusion feature space R
nCompare with existing signal characteristic space method for building up, the feature space that the inventive method is set up can make target echo and reverberation have stable more and pervasive separation property.Achievement of the present invention is not limited in the application of water-bed Target Recognition; Can also be widely used in Forward-looking Sonar, unmanned active sonar Target Recognition fields such as device, frogman's sonar of diving under water; Improve the detection performance of active sonar; The realization telesecurity is surveyed, for the offshore defensive sonar provides new implementation.
Description of drawings
Fig. 1 is associating modeling of the present invention and recognition methods performing step synoptic diagram.
Embodiment
For example the present invention is done more detailed description below in conjunction with accompanying drawing:
In conjunction with Fig. 1.Be transformed to example with Fourier Transform of Fractional Order and Hilbert-Huang, content described herein does not play the qualification effect to content of the present invention.The present invention includes following steps:
Step 1:
According to the highlight model of target echo signal, when active sonar transmitted to linear FM signal, target echo was a linear FM signal also, had energy accumulating characteristic and many component characteristics.According to these characteristics, can adopt Fourier Transform of Fractional Order and Hilbert-Huang transfer pair echoed signal to carry out feature extraction, extract energy accumulating property characteristic and many minutes measure features of echoed signal respectively.
For echoed signal x (t), it is carried out suc as formula the Fourier Transform of Fractional Order shown in (1).
Wherein, kernel function K
p(u, t) be one with time t, fractional order territory coordinate axis u and the relevant function of fractional order power p.Through changing the value of p, make f
p(u) it is maximum that peak value reaches, at this moment f
p(u) be the energy accumulating property characteristic of echoed signal.
The Hilbert-Huang conversion of echoed signal x (t) is x (t) to be carried out EMD decompose; Envelope to signal carries out match; Set up the establishment condition of IMF component; Employing deducts from signal up and down that the mode of envelope mean value obtains each rank IMF component step by step, at last with signal decomposition be several IMF components and a remainder and.Give up remainder, all the other IMF components are done the Hilbert conversion, can obtain the Hilbert spectrum of echoed signal.
Through above two kinds of signal characteristic extracting methods, two feature spaces in the step 1 among Fig. 1 have promptly been set up.
Step 2: two steps of dimensionality reduction and fusion that comprised the signal characteristic space.
Through step 1, echoed signal x (t) has been mapped to a certain specific signal feature space from time domain.This signal characteristic space is compressed, and a principle should following is in the signal characteristic space after compression, and separability is higher between the class of target echo and reverberation.Therefore; Under the Euclidean distance meaning, adopt suc as formula the compression criterion of the Fisher criterion function shown in (2) as the signal characteristic space, purpose is in the signal characteristic space after compression; Two types of sample of signal have distribution distance in the littler class, and scatter distance between bigger class.
Wherein, J
FBe the objective function of separability between two types of class signals of reflection, w is used for the signal characteristic space is carried out the projecting direction vector of projection dimensionality reduction, S
wBe scatter matrix in the class of sample of signal in the feature space, S
bBe scatter matrix between the class of sample of signal in the feature space.Adopt method of Lagrange multipliers that formula (2) is found the solution, try to achieve one group of projecting direction vector that satisfies orthogonality condition.
With the signal characteristic space of setting up in the step 1
On this group vector, carry out projection, just obtained new feature space, be defined as R '
i, shown in step 2 in the accompanying drawing.R '
iSpatial Dimension to be lower than the original signal feature space
So just realized the dimensionality reduction in signal characteristic space.
For two signal characteristic spaces that obtain through step 1
With
Adopt above-mentioned feature compression method to obtain two low dimensional feature space R ' respectively
1And R '
2For the information with these two signal characteristic spaces merges, adopt canonical correlation analysis to extract the canonical correlation variable between the eigenvector of representing same sample of signal in two signal characteristic spaces.The canonical correlation variable of the eigenvector of the same sample of signal of representative in two signal characteristic spaces is carried out end to end splicing, just obtain a new fusion feature space, be defined as R
nIn this example, n=2d.
Step 3:
The R in the fusion feature space
nIn, according to target echo and the distribution of reverberation in feature space, can between two types of sample of signal, set up a classifying face g (z) who describes with math equation, be defined as based on the decision rule of this classifying face,
The form of g (z) is the linear weighted function combination of each dimension of feature space, and each weights can be obtained through optimization method according to known sample.Make that g (z) is zero, what obtain is exactly the math equation of classifying face between two types of signals in the feature space.According to this equation, promptly can set up target echo and reverberation at feature space R
nIn distributed model.
Step 4:
For the echo signal samples of new acquisition, adopt step 1 it to be carried out signal characteristic and extract, and obtain its eigenvector z in the fusion feature space with the said method of step 2, with its substitution decision rule,, just can judge the affiliated classification of z according to decision rule.
Claims (5)
1. R
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation is characterized in that may further comprise the steps:
Step 1: with reverberation as one type of signal source with stabilization signal character; Foundation comprises the training sample database of target echo and reverberation; Based on two kinds of signal differences of physical properties, adopt the unlike signal processing method that training sample is mapped in a plurality of signal characteristics space;
Step 2:,, and, set up and merge the signal characteristic space to the fusion of various signals feature space to signal characteristic space dimensionality reduction according to the distributed intelligence of training sample in the signal characteristic space;
Step 3: set up target echo and the joint distribution model of reverberation in merging the signal characteristic space in the training sample;
Step 4: for the sample of signal of the unknown, be mapped in the fusion feature space, and compare, judge the kind of unknown signaling sample with joint distribution model that step 3 is set up through step 1 and the described signal processing method of step 2.
2. a kind of R according to claim 1
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation; It is characterized in that: described employing unlike signal disposal route is meant and adopts Fourier Transform of Fractional Order that training sample is mapped to energy accumulating property feature space and adopts the Hilbert-Huang conversion that training sample is mapped to many components feature space.
3. a kind of R according to claim 1
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation is characterized in that: described signal characteristic space dimensionality reduction, employing be under the Euclidean distance meaning, carry out based on the linear discriminant analysis of Fisher criterion function.
4. a kind of R according to claim 1
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation is characterized in that: merge in described signal characteristic space, employing be that canonical correlation analysis combines the method for series connection Feature Fusion to carry out.
5. a kind of R according to claim 1
nThe water-bed target echo in the space and the characteristic binding modeling method of reverberation is characterized in that: the modeling of described feature space joint distribution, employing be that the discriminant function method is set up math equation to target echo and the classifying face of reverberation in the fusion feature space.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101500725A CN102692625A (en) | 2012-05-15 | 2012-05-15 | Joint modeling method for features of underwater target echo and reverberation in Rn space |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101500725A CN102692625A (en) | 2012-05-15 | 2012-05-15 | Joint modeling method for features of underwater target echo and reverberation in Rn space |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102692625A true CN102692625A (en) | 2012-09-26 |
Family
ID=46858208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012101500725A Pending CN102692625A (en) | 2012-05-15 | 2012-05-15 | Joint modeling method for features of underwater target echo and reverberation in Rn space |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102692625A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015172622A1 (en) * | 2014-05-14 | 2015-11-19 | 武汉大学 | Method for radio-frequency interference suppression of high-frequency ground wave radar |
| CN110070031A (en) * | 2019-04-18 | 2019-07-30 | 哈尔滨工程大学 | A kind of sediment extracting echo characteristics of active sonar fusion method based on EMD and random forest |
| CN110850421A (en) * | 2019-11-21 | 2020-02-28 | 中国科学院声学研究所 | Underwater target detection method based on space-time adaptive processing of reverberation symmetric spectrum |
| CN111931803A (en) * | 2020-06-17 | 2020-11-13 | 中国船舶重工集团公司第七一五研究所 | Evaluation criterion method for fine features of underwater acoustic signals |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080175434A1 (en) * | 2007-01-18 | 2008-07-24 | Northrop Grumman Systems Corporation | Automatic target recognition system for detection and classification of objects in water |
| CN101470194A (en) * | 2007-12-26 | 2009-07-01 | 中国科学院声学研究所 | Torpedo target recognition method |
| CN101900810A (en) * | 2010-07-15 | 2010-12-01 | 哈尔滨工程大学 | Method for fusing multi-probe end sonar information by using submersible as carrier |
| CN102142136A (en) * | 2011-03-05 | 2011-08-03 | 河海大学常州校区 | Neural network based sonar image super-resolution reconstruction method |
-
2012
- 2012-05-15 CN CN2012101500725A patent/CN102692625A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080175434A1 (en) * | 2007-01-18 | 2008-07-24 | Northrop Grumman Systems Corporation | Automatic target recognition system for detection and classification of objects in water |
| CN101470194A (en) * | 2007-12-26 | 2009-07-01 | 中国科学院声学研究所 | Torpedo target recognition method |
| CN101900810A (en) * | 2010-07-15 | 2010-12-01 | 哈尔滨工程大学 | Method for fusing multi-probe end sonar information by using submersible as carrier |
| CN102142136A (en) * | 2011-03-05 | 2011-08-03 | 河海大学常州校区 | Neural network based sonar image super-resolution reconstruction method |
Non-Patent Citations (3)
| Title |
|---|
| LI TINGTING等: "Classification of Underwater Mines by Means of the FRFT and SVM", 《PROCEEDINGS OF THE 2010 IEEE INTERNATIONAL CONFERENCE ON INFORMATION AND AUTOMATION》 * |
| 李秀坤等: "水下目标特性特征提取及其融合", 《哈尔滨工程大学学报》 * |
| 谢磊等: "水雷目标的亮点信息提取及融合研究", 《声学技术》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015172622A1 (en) * | 2014-05-14 | 2015-11-19 | 武汉大学 | Method for radio-frequency interference suppression of high-frequency ground wave radar |
| CN110070031A (en) * | 2019-04-18 | 2019-07-30 | 哈尔滨工程大学 | A kind of sediment extracting echo characteristics of active sonar fusion method based on EMD and random forest |
| CN110850421A (en) * | 2019-11-21 | 2020-02-28 | 中国科学院声学研究所 | Underwater target detection method based on space-time adaptive processing of reverberation symmetric spectrum |
| CN111931803A (en) * | 2020-06-17 | 2020-11-13 | 中国船舶重工集团公司第七一五研究所 | Evaluation criterion method for fine features of underwater acoustic signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106066468B (en) | It is a kind of based on acoustic pressure, the vector array port/starboard discrimination method of vibration velocity Mutual spectrum | |
| Sasaki et al. | Three-dimensional imaging method incorporating range points migration and Doppler velocity estimation for UWB millimeter-wave radar | |
| CN102692625A (en) | Joint modeling method for features of underwater target echo and reverberation in Rn space | |
| CN102279390A (en) | Intra-pulse modulation and recognition method of low signal-to-noise radar radiation source signal | |
| CN104330787B (en) | Underwater motion array multi-target detection and position estimation integrated method | |
| CN103941244A (en) | Radar target one-dimensional range profile local optimal sub-space recognition method | |
| CN106127110A (en) | A kind of human body fine granularity motion recognition method based on UWB radar with optimum SVM | |
| CN105974376A (en) | SAR radio frequency interference suppressing method | |
| CN102636775B (en) | Wind profile radar echo spectrum reconfiguration method based on fuzzy logic recognition | |
| CN106814360B (en) | A kind of multibeam sounding system based on linear FM signal | |
| CN105425223A (en) | Detection method of sparse distance extension radar target in generalized Pareto clutter | |
| Murphy et al. | Examining the robustness of automated aural classification of active sonar echoes | |
| CN104714237A (en) | Fish identification method with multi-feature and multidirectional data fused | |
| CN103267964A (en) | Missile-borne seeker radar Sigma-Delta-STAP method based on low-rank matrix recovery | |
| CN103487796B (en) | A kind of method utilizing underwater acoustic channel Statistically invariant feature to realize passive ranging | |
| US20190146054A1 (en) | Wave source direction estimation apparatus, wave source direction estimation system, wave source direction estimation method, and wave source direction estimation program | |
| Kubicek et al. | Sonar target representation using two-dimensional Gabor wavelet features | |
| CN104062663A (en) | Multi-beam seafloor sub-bottom profile detection device | |
| CN106526577B (en) | A kind of array shape estimation method using cooperation sound source information | |
| CN104665875A (en) | Ultrasonic Doppler envelope and heart rate detection method | |
| CN101576618A (en) | Acoustic positioning measurement method based on wavelet transformation and measurement system thereof | |
| CN103926581A (en) | Sonar target echo highlight parameter measurement method | |
| CN103617628A (en) | Naval ship detection method based on polarized azimuth secant function characteristic | |
| Dawood et al. | Superresolution Doppler estimation using UWB random noise signals and MUSIC | |
| CN113242197A (en) | Underwater acoustic signal modulation identification method and system based on artificial intelligence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120926 |