CN110850420B - Fisher SVM sonar signal discrimination method based on marble loss - Google Patents

Fisher SVM sonar signal discrimination method based on marble loss Download PDF

Info

Publication number
CN110850420B
CN110850420B CN201911121646.4A CN201911121646A CN110850420B CN 110850420 B CN110850420 B CN 110850420B CN 201911121646 A CN201911121646 A CN 201911121646A CN 110850420 B CN110850420 B CN 110850420B
Authority
CN
China
Prior art keywords
model
discrimination
sonar signal
sample set
result
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.)
Active
Application number
CN201911121646.4A
Other languages
Chinese (zh)
Other versions
CN110850420A (en
Inventor
张莉
张正齐
王邦军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou University
Original Assignee
Suzhou University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou University filed Critical Suzhou University
Priority to CN201911121646.4A priority Critical patent/CN110850420B/en
Publication of CN110850420A publication Critical patent/CN110850420A/en
Application granted granted Critical
Publication of CN110850420B publication Critical patent/CN110850420B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
    • G01S15/88—Sonar systems specially adapted for specific applications
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/52—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S15/00
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Acoustics & Sound (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

The application discloses a Fisher SVM sonar signal discrimination method, device, equipment and readable storage medium based on marble loss, and the scheme comprises the following steps: constructing a discrimination model based on a Fisher criterion regularization support vector machine; optimizing the discrimination model by using the sample set according to the marble loss function to determine a weight vector and a deviation value of the discrimination model; inputting sonar signal data of the target detection object into the discrimination model to obtain a model output result, and determining the type discrimination result of the target detection object according to the model output result. Therefore, the scheme adopts the marble loss function to optimize the discrimination model, and due to the use of the marble loss function, the anti-noise performance of the discrimination model is improved, so that the sonar signal discrimination scheme has good classification performance and stability.

Description

Fisher SVM sonar signal discrimination method based on marble loss
Technical Field
The application relates to the technical field of computers, in particular to a Fisher SVM sonar signal distinguishing method, device and equipment based on marble loss and a readable storage medium.
Background
With the rapid development of science and technology, the understanding of the oceans covering 70% of the total area of the earth is deepened, and the oceans are more and more emphasized by people due to the huge economic potential and the strategic importance.
Due to the special properties of seawater, electromagnetic waves and light waves cannot effectively transmit information in seawater. Experiments prove that the propagation performance of sound waves in seawater is the best among various well-known radiation signals. The attenuation of the sound wave propagating in water is relatively small, and the low-frequency sound wave can penetrate through the stratum of the seabed even by thousands of meters. Therefore, both submarines and surface ships can only detect underwater objects by using a sonar system. A reliable sonar signal distinguishing system can quickly distinguish whether a detected object is a common rock or a metal object and quickly and accurately position and detect metal resources.
A Support Vector Machine (SVM) established on the basis of a VC Dimension (Vapnik Chervonenks Dimension) theory of statistics and a principle of minimizing structural risk can be applied to pattern recognition and regression analysis. The relativistic scholars combine linear discriminant analysis and SVM to propose a Fisher regularization support vector machine (Fisher SVM) from the perspective of regularization. For the binary classification problem, Fisher SVM can find a classification hyperplane and simultaneously maximize the inter-class spacing and minimize the intra-class divergence. Briefly, Fisher SVM can approximately satisfy the Fisher criterion and achieve good statistical separability. However, both the traditional SVM and the FisherSVM are based on the Hinge loss function, and the Hinge loss function has a certain defect in the anti-noise performance, so that the sonar signal discrimination scheme based on the FisherSVM has a poor discrimination effect.
Disclosure of Invention
The application aims to provide a Fisher SVM sonar signal discrimination method, device, equipment and readable storage medium based on marble loss, and aims to solve the problem that the conventional Fisher SVM sonar signal discrimination scheme is poor in anti-noise performance and low in discrimination result reliability. The specific scheme is as follows:
in a first aspect, the application provides a FisherSVM sonar signal discrimination method based on pinball loss, which includes:
constructing a discrimination model based on a Fisher criterion regularization support vector machine;
optimizing the discrimination model by utilizing a sample set according to a marble loss function to determine a weight vector and a deviation value of the discrimination model;
inputting sonar signal data of the target detection object into the discrimination model to obtain a model output result, and determining the type discrimination result of the target detection object according to the model output result.
Preferably, the inputting sonar signal data of the target detection object into the discrimination model includes:
and carrying out normalization operation on the sonar signal data of the target detection object according to a preset normalization mode, and inputting the normalized sonar signal data into the discrimination model.
Preferably, the sonar signal data includes a plurality of characteristics, and each of the characteristics is used for representing energy in a corresponding frequency band within a preset time.
Preferably, the sample set comprises a plurality of sets of sonar signal data with class labels, wherein the class labels comprise rock and metal.
Preferably, the determining the type discrimination result of the target detection object according to the model output result includes:
obtaining a type discrimination result of the target detection object according to the model output result and a preset formula, wherein the preset formula is as follows:
Figure BDA0002275607920000021
wherein the content of the first and second substances,
Figure BDA0002275607920000022
sign () is a sign function for a model output result of the discriminant model; if it is
Figure BDA0002275607920000023
If the number is 1, the type discrimination result is metal, if so, the type discrimination result is metal
Figure BDA0002275607920000024
And if the result is-1, the type discrimination result is rock.
Preferably, the constructing a discriminant model based on a Fisher criterion regularization support vector machine includes:
constructing a discrimination model based on a Fisher criterion regularization support vector machine, wherein the discrimination function of the discrimination model is as follows:
Figure BDA0002275607920000031
wherein the content of the first and second substances,
Figure BDA0002275607920000032
outputting a result for the model of the discriminant model, wherein l represents the number of samples in the sample set, xiRepresents the ith sample, a, in the set of samplesiDenotes xiX represents the input vector, k (-) is the kernel function, and b represents the deviation value.
Preferably, the optimizing the discriminant model by using a sample set according to a pinball loss function to determine a weight vector and a bias value of the discriminant model includes:
determining an objective optimization function of the discriminant model, wherein the objective optimization function is as follows:
Figure BDA0002275607920000033
s.t.Dy(Kα+b1)≥1-ξ,
Figure BDA0002275607920000034
wherein, alpha represents a weight vector, and b represents an offset value; xi is a relaxation variable; l represents the number of samples of the sample set; gamma rayKAnd gammaFIs a non-negative regularization parameter; k is a kernel matrix; n is equal to I-G,
Figure BDA0002275607920000035
is a unit matrix, G is a preset matrix; dyIs a diagonal matrix, and DyRow i and column i of (2)y]ii＝yi，i＝1,2,...,l，yiA category label for the ith sample in the sample set;
performing form conversion on the relaxation variables in the target optimization function according to a marble loss function, and optimizing the discriminant model by using a sample set to determine a weight vector and a deviation value of the discriminant model, wherein the relaxation variables after conversion
Figure BDA0002275607920000036
The pinball loss function is:
Figure BDA0002275607920000037
wherein tau is more than 0 and less than or equal to 1.
In a second aspect, the present application provides a fisher svm sonar signal discriminating device based on pinball loss, including:
a model construction module: the method comprises the steps of constructing a discrimination model based on a Fisher criterion regularization support vector machine;
a model optimization module: the system comprises a sample set, a weight vector and a deviation value, wherein the sample set is used for optimizing the discriminant model according to a marble loss function so as to determine the weight vector and the deviation value of the discriminant model;
a result judging module: and the sonar signal data of the target detection object is input into the discrimination model to obtain a model output result, and the type discrimination result of the target detection object is determined according to the model output result.
In a third aspect, the present application further provides a fisher svm sonar signal discriminating device based on pinball loss, including:
a memory: for storing a computer program;
a processor: for executing the computer program to realize the steps of the FisherSVM sonar signal discrimination method based on pinball loss as described above.
In a fourth aspect, the present application further provides a readable storage medium, on which a computer program is stored, the computer program, when being executed by a processor, is used for implementing the steps of the FisherSVM sonar signal discrimination method based on pinball loss as described above.
The application provides a Fisher SVM sonar signal discrimination method based on marble loss, which comprises the following steps: constructing a discrimination model based on a Fisher criterion regularization support vector machine; optimizing the discrimination model by using the sample set according to the marble loss function to determine a weight vector and a deviation value of the discrimination model; inputting sonar signal data of the target detection object into the discrimination model to obtain a model output result, and determining the type discrimination result of the target detection object according to the model output result. Therefore, the method optimizes the discrimination model by adopting the pinball loss function, and due to the use of the pinball loss function, the anti-noise performance of the discrimination model is improved, so that the sonar signal discrimination scheme has good classification performance and stability.
In addition, the application also provides a Fisher SVM sonar signal distinguishing device, equipment and a readable storage medium based on marble loss, and the technical effect of the method corresponds to that of the method, and the method is not repeated herein.
Drawings
For a clearer explanation of the embodiments or technical solutions of the prior art of the present application, the drawings needed for the description of the embodiments or prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flow chart of an implementation of a FisherSVM sonar signal discrimination method based on pinball loss according to an embodiment of the present application;
fig. 2 is a flow chart of an implementation of a FisherSVM sonar signal discrimination method based on pinball loss according to a second embodiment of the present disclosure;
fig. 3 is a functional block diagram of an embodiment of a FisherSVM sonar signal discrimination device based on pinball loss according to the present application;
fig. 4 is a schematic structural diagram of an embodiment of FisherSVM sonar signal discrimination equipment based on marble loss according to the present application.
Detailed Description
The core of the application is to provide a Fisher SVM sonar signal discrimination method, device, equipment and readable storage medium based on pinball loss, a pinball loss function is adopted to optimize a discrimination model, the anti-noise performance of the discrimination model is improved, and the sonar signal discrimination scheme has good classification performance and stability.
In order that those skilled in the art will better understand the disclosure, the following detailed description will be given with reference to the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, a first embodiment of a FisherSVM sonar signal discrimination method based on pinball loss provided by the present application is described below, where the first embodiment includes:
s101, constructing a discrimination model based on a Fisher criterion regularization support vector machine;
s102, optimizing the discriminant model by utilizing a sample set according to a marble loss function to determine a weight vector and a deviation value of the discriminant model;
s103, inputting sonar signal data of the target detection object into the judgment model to obtain a model output result, and determining the type judgment result of the target detection object according to the model output result.
In the implementation process, firstly, bounce sonar signals of the target detection object are collected, the sonar signals comprise a plurality of characteristics, each characteristic represents energy in a specific frequency band within a period of time, and then the classification of the target detection object according to sonar signal data is realized by applying the embodiment. The embodiment can be applied to various scenes, for example, in the ocean detection process, the embodiment can be applied to judge whether the type of the target detection object is metal or rock.
In the implementation process of this embodiment, a discriminant model based on a Fisher criterion regularization support vector machine, i.e., the Fisher regularization support vector machine, is first constructed. After the discriminant model is obtained, the discriminant model is further optimized in the embodiment, and particularly, the discriminant model is optimized by using a pinball loss function, so that a weight vector and a deviation value after the discriminant model is optimized are obtained. In addition, a sample set is needed in the optimization process, and the sample set comprises a plurality of sonar signal data with category labels. Finally, the sonar signal data of the target detection object can be input into the optimized discrimination model to obtain a model output result, and the type of the target detection object is determined according to the model output result.
In summary, the FisherSVM sonar signal discriminating method based on pinball loss provided by the embodiment includes: constructing a discrimination model based on a Fisher criterion regularization support vector machine; optimizing the discrimination model by using the sample set according to the marble loss function to determine a weight vector and a deviation value of the discrimination model; inputting sonar signal data of the target detection object into the discrimination model to obtain a model output result, and determining the type discrimination result of the target detection object according to the model output result. Therefore, the method optimizes the discrimination model by adopting the pinball loss function, and due to the use of the pinball loss function, the anti-noise performance of the discrimination model is improved, so that the sonar signal discrimination scheme has good classification performance and stability.
The second embodiment of the FisherSVM sonar signal discrimination method based on pinball loss provided by the present application is described in detail below, and is implemented based on the first embodiment, and is expanded to a certain extent on the basis of the first embodiment.
Referring to fig. 2, the second embodiment specifically includes:
s201, constructing a sample set;
the sample set comprises a plurality of sets of sonar signal data with category labels, wherein the category labels comprise rocks and metals, and the sonar signal data comprises a plurality of characteristics, and each characteristic is used for representing energy in a corresponding frequency band within a preset time.
Specifically, relevant data of the ocean sonar detection is collected statistically as a sample set in the embodiment. Set of samples as
Figure BDA0002275607920000071
Wherein
Figure BDA0002275607920000072
yiE { + -1 }. Wherein the label is yi1 is a set of sonar signal data for detecting metal objects as objects, and the label is yiAnd 1, the feature number of each sample is d, and l is the total number of samples in the training set. Normalizing the data in the sample set S to obtain a sampleThis point is mapped to the interval [0,1]]In (1).
S202, constructing a discrimination model based on a Fisher criterion regularization support vector machine;
specifically, in this embodiment, the discriminant function of the discriminant model is as follows:
Figure BDA0002275607920000073
wherein the content of the first and second substances,
Figure BDA0002275607920000074
outputting a result for the model of the discriminant model, wherein l represents the number of samples in the sample set, xiRepresents the ith sample, a, in the set of samplesiDenotes xiX represents the input vector, k (-) is the kernel function, and b represents the deviation value.
S203, determining a target optimization function of the discriminant model;
wherein the objective optimization function is:
Figure BDA0002275607920000075
s.t.Dy(Kα+b1)≥1-ξ,
Figure BDA0002275607920000076
α represents a weight vector, and b represents a bias value; xi is a relaxation variable; l represents the number of samples of the sample set; gamma rayKAnd gammaFIs a non-negative regularization parameter; k is a kernel matrix; n is equal to I-G,
Figure BDA0002275607920000077
is a unit matrix, G is a preset matrix; dyIs a diagonal matrix, and DyRow i and column i of (2)y]ii＝yi，i＝1,2,...,l，yiFor the ith in the sample setA category label for the sample;
s204, performing form conversion on a relaxation variable in the target optimization function according to a marble loss function, and optimizing the discriminant model by using a sample set to determine a weight vector and a deviation value of the discriminant model;
in particular, the relaxation variables after conversion
Figure BDA0002275607920000081
In particular, in the above discriminant model
Figure BDA0002275607920000082
Is a general sample, not specifically a sample, which corresponds to the second variable k (x) in the kernel functioniX). And herein is
Figure BDA0002275607920000083
X in (1) is subscriptable, which in this case refers to a particular sample. Each sample in the constraint, which is a specific sample set, is to be evaluated and substituted into the functional expression of the discriminant model.
The pinball loss function is:
Figure BDA0002275607920000084
0＜τ≤1。
s205, carrying out normalization operation on sonar signal data of the target detection object according to a preset normalization mode, and inputting the normalized sonar signal data into the discrimination model to obtain a model output result;
the preset normalization mode refers to the same normalization mode as that in the process of constructing the sample set in S201, and a specific normalization mode may be selected according to actual requirements, which is not specifically limited in this embodiment.
And S206, obtaining a type discrimination result of the target detection object according to the model output result and a preset formula.
Wherein the preset formula is as follows:
Figure BDA0002275607920000085
wherein the content of the first and second substances,
Figure BDA0002275607920000086
sign () is a sign function for a model output result of the discriminant model; if it is
Figure BDA0002275607920000087
If the number is 1, the type discrimination result is metal, if so, the type discrimination result is metal
Figure BDA0002275607920000088
And if the result is-1, the type discrimination result is rock.
Therefore, the FisherSVM sonar signal discrimination method based on pinball loss provided by the embodiment mainly comprises a data preprocessing process, a model optimization process and a data prediction process, wherein in the data preprocessing process, category labels are added to sonar signal data, and normalization is performed to obtain a sample set; in the model optimization process, optimizing the discrimination model by using a sample set according to a marble loss function so as to determine a weight vector and a bias value of the discrimination model; in the data prediction process, the optimized discrimination model is used for classifying the sonar signal data of the target detection object, and finally the type of the target detection object is determined.
Based on the embodiment of the FisherSVM sonar signal discrimination method based on the marble loss, the following application takes a specific application scenario as an example to introduce the specific implementation process and implementation effect of the scheme.
In particular, the present application tests on the Sonar dataset from the UCI, which classifies target objects according to the bouncing Sonar signals collected through various angles and under various conditions. The data set contains a total of 208 samples, each sample containing 60 features, each feature representing energy within a particular frequency band over a period of time. Wherein the number of the rock sample and the number of the metal sample are 97 and 111 respectively, and the label of the metal sample is marked as +1 and the label of the rock sample is marked as-1.
In order to prove that the embodiment of the application is insensitive to noise interference, and has good classification performance and stability for noise-disturbed samples. Here, noise interference contrast is added, the same gaussian noise is covered on the whole data set, and the variance ratio of the noise to the original data is r. The present application covers gaussian noise with r ═ {0, 0.05, 0.1} for all original samples (where r ═ 0 is the original sample), and performs the following steps for one original sample and two gaussian noise samples, respectively:
first, data preprocessing part
Inputting Sonar sample set
Figure BDA0002275607920000091
Wherein
Figure BDA0002275607920000092
yiE { + -1 }. The label is yi1 is a set of sonar signal data for detecting metal objects as objects, and the label is yiThe data set of sonar signals is-1, the feature number of each sample is 60, and the total number of samples is 208. 2/3 of the sample set were randomly chosen as the training set, and the remainder 1/3 was the test set.
Second, model optimization part
The application constructs a discriminant function as follows:
Figure BDA0002275607920000093
where α is the weight vector of the function and b is the deviation of the function. To obtain the function weight vector and the deviation, the following optimization problem is required to be solved:
Figure BDA0002275607920000101
s.t.Dy(Kα+b1)≥1-ξ,
Figure BDA0002275607920000102
wherein, alpha represents a weight vector, and b represents an offset value; xi is a relaxation variable; l represents the number of samples of the sample set; gamma rayKAnd gammaFIs a non-negative regularization parameter; k is a kernel matrix; n is equal to I-G,
Figure BDA0002275607920000103
is a unit matrix, G is a preset matrix; dyIs a diagonal matrix, and DyRow i and column i of (2)y]ii＝yi，i＝1,2,...,l，yiA category label for the ith sample in the sample set; relaxation variables
Figure BDA0002275607920000104
Lτ(. cndot.) is defined as the marble loss function as follows:
Figure BDA0002275607920000105
where 0 < τ ≦ 1, τ is here given as {0.1, 0.3, 0.5, 0.6, 0.8, 1 }.
After solving the above optimization problem, α and b are obtained, so that the discriminant function can be determined.
Data prediction part
Inputting customer data x to be predicted, mapping the customer data x into an interval [0,1] according to a data normalization mode in a data preprocessing module, and then substituting a discriminant function to calculate an estimated value:
Figure BDA0002275607920000106
and then judging the sonar signals returned by the detected object according to the following rules:
Figure BDA0002275607920000107
if it is
Figure BDA0002275607920000108
And 1, the detected object is a metal object, otherwise, the detected object is a rock.
Table 1 lists the accuracy of the present application compared to FisherSVM at certain parameter settings. It can be seen that under different parameter settings, the embodiments of the present application have better results than FisherSVM.
TABLE 1
Figure BDA0002275607920000109
Figure BDA0002275607920000111
In the following, a FisherSVM sonar signal discriminating device based on pinball loss according to an embodiment of the present application is introduced, and a FisherSVM sonar signal discriminating device based on pinball loss described below and a FisherSVM sonar signal discriminating method based on pinball loss described above may be referred to in correspondence.
As shown in fig. 3, the apparatus embodiment comprises:
model building module 301: the method comprises the steps of constructing a discrimination model based on a Fisher criterion regularization support vector machine;
the model optimization module 302: the system comprises a sample set, a weight vector and a deviation value, wherein the sample set is used for optimizing the discriminant model according to a marble loss function so as to determine the weight vector and the deviation value of the discriminant model;
the discrimination result module 303: and the sonar signal data of the target detection object is input into the discrimination model to obtain a model output result, and the type discrimination result of the target detection object is determined according to the model output result.
The fisher svm sonar signal discriminating device based on pinball loss in this embodiment is used for implementing the fisher svm sonar signal discriminating method based on pinball loss, and therefore the specific implementation manner in the device can be seen in the foregoing embodiments of the fisher svm sonar signal discriminating method based on pinball loss, for example, the model constructing module 301, the model optimizing module 302, and the discrimination result module 303 are respectively used for implementing steps S101, S102, and S103 in the fisher svm sonar signal discriminating method based on pinball loss. Therefore, specific embodiments thereof may be referred to in the description of the corresponding respective partial embodiments, and will not be described herein.
In addition, since the FisherSVM sonar signal discriminating device based on pinball loss of the embodiment is used for realizing the FisherSVM sonar signal discriminating method based on pinball loss, the function corresponds to that of the method, and details are not repeated here.
In addition, the present application also provides a FisherSVM sonar signal discriminating device based on pinball loss, as shown in fig. 4, including:
the memory 100: for storing a computer program;
the processor 200: for executing the computer program to realize the steps of the FisherSVM sonar signal discrimination method based on pinball loss as described above.
Finally, the present application provides a readable storage medium having stored thereon a computer program for implementing, when being executed by a processor, the steps of a FisherSVM sonar signal discrimination method based on pinball loss as described above.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The above detailed descriptions of the solutions provided in the present application, and the specific examples applied herein are set forth to explain the principles and implementations of the present application, and the above descriptions of the examples are only used to help understand the method and its core ideas of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims (7)

1. A FisherSVM sonar signal distinguishing method based on marble loss is characterized by comprising the following steps:
constructing a discrimination model based on a Fisher criterion regularization support vector machine;
optimizing the discrimination model by utilizing a sample set according to a marble loss function so as to determine a weight vector and a deviation value of the discrimination model, wherein the sample set comprises sonar signal data of multiple sets of band class labels, and the class labels comprise rocks and metals;
inputting sonar signal data of a target detection object into the discrimination model to obtain a model output result, and determining the type discrimination result of the target detection object according to the model output result, wherein the sonar signal data comprise a plurality of characteristics, and each characteristic is used for representing energy in a corresponding frequency band within preset time;
the optimizing the discriminant model by using a sample set according to a pinball loss function to determine a weight vector and a deviation value of the discriminant model includes:
determining an objective optimization function of the discriminant model, wherein the objective optimization function is as follows:
Figure FDA0003270320270000011
s.t.Dy(Kα+b1)≥1-ξ,
Figure FDA0003270320270000012
wherein, alpha represents a weight vector, and b represents an offset value; xi is a relaxation variable; l represents the number of samples of the sample set; gamma rayKAnd gammaFIs a non-negative regularization parameter; k is a kernel matrix; n is equal to I-G,
Figure FDA0003270320270000013
is a unit matrix, G is a preset matrix; dyIs a diagonal matrix, and DyRow i and column i of (2)y]ii＝yi，i＝1,2,...,l，yiA category label for the ith sample in the sample set;
performing form conversion on the relaxation variables in the target optimization function according to a marble loss function, and optimizing the discriminant model by using a sample set to determine a weight vector and a deviation value of the discriminant model, wherein the relaxation variables after conversion
Figure FDA0003270320270000014
The pinball loss function is:
Figure FDA0003270320270000015
wherein tau is more than 0 and less than or equal to 1.
2. The method according to claim 1, wherein inputting sonar signal data for a target detection object into the discriminative model comprises:
and carrying out normalization operation on the sonar signal data of the target detection object according to a preset normalization mode, and inputting the normalized sonar signal data into the discrimination model.
3. The method of claim 1, wherein said determining a type discrimination result for the target detection object based on the model output result comprises:
obtaining a type discrimination result of the target detection object according to the model output result and a preset formula, wherein the preset formula is as follows:
Figure FDA0003270320270000021
wherein the content of the first and second substances,
Figure FDA0003270320270000022
sign () is a sign function for a model output result of the discriminant model; if it is
Figure FDA0003270320270000023
If the number is 1, the type discrimination result is metal, if so, the type discrimination result is metal
Figure FDA0003270320270000024
And if the result is-1, the type discrimination result is rock.
4. The method of claim 1, wherein constructing a discriminant model based on a Fisher criterion regularization support vector machine comprises:
constructing a discrimination model based on a Fisher criterion regularization support vector machine, wherein the discrimination function of the discrimination model is as follows:
Figure FDA0003270320270000025
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA0003270320270000026
outputting a result for the model of the discriminant model, wherein l represents the number of samples in the sample set, xiRepresents the ith sample, a, in the set of samplesiDenotes xiX represents the input vector, k (-) is the kernel function, and b represents the deviation value.
5. A FisherSVM sonar signal discriminating device based on marble loss is characterized by comprising:
a model construction module: the method comprises the steps of constructing a discrimination model based on a Fisher criterion regularization support vector machine;
a model optimization module: the system comprises a sample set, a judgment model and a database, wherein the sample set is used for optimizing the judgment model according to a marble loss function so as to determine a weight vector and a deviation value of the judgment model, and comprises sonar signal data with multiple band class labels, wherein the class labels comprise rocks and metals;
a result judging module: the sonar signal data comprises a plurality of characteristics, and each characteristic is used for representing the energy in a corresponding frequency band within preset time;
the model optimization module is specifically configured to:
determining an objective optimization function of the discriminant model, wherein the objective optimization function is as follows:
Figure FDA0003270320270000027
s.t.Dy(Kα+b1)≥1-ξ,
Figure FDA0003270320270000028
wherein, alpha represents a weight vector, and b represents an offset value; xi is a relaxation variable; l meterIndicating the number of samples of the sample set; gamma rayKAnd gammaFIs a non-negative regularization parameter; k is a kernel matrix; n is equal to I-G,
Figure FDA0003270320270000031
is a unit matrix, G is a preset matrix; dyIs a diagonal matrix, and DyRow i and column i of (2)y]ii＝yi，i＝1,2,...,l，yiA category label for the ith sample in the sample set;
performing form conversion on the relaxation variables in the target optimization function according to a marble loss function, and optimizing the discriminant model by using a sample set to determine a weight vector and a deviation value of the discriminant model, wherein the relaxation variables after conversion
Figure FDA0003270320270000032
The pinball loss function is:
Figure FDA0003270320270000033
wherein tau is more than 0 and less than or equal to 1.
6. A FisherSVM sonar signal discrimination device based on marble loss is characterized by comprising:
a memory: for storing a computer program;
a processor: for executing said computer program to implement the steps of a FisherSVM sonar signal discrimination method based on pinball loss as claimed in any one of claims 1 to 4.
7. A readable storage medium, having stored thereon a computer program for implementing the steps of the method for fisher svm sonar signal discrimination based on pinball loss according to any one of claims 1 to 4 when executed by a processor.
CN201911121646.4A 2019-11-15 2019-11-15 Fisher SVM sonar signal discrimination method based on marble loss Active CN110850420B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911121646.4A CN110850420B (en) 2019-11-15 2019-11-15 Fisher SVM sonar signal discrimination method based on marble loss

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911121646.4A CN110850420B (en) 2019-11-15 2019-11-15 Fisher SVM sonar signal discrimination method based on marble loss

Publications (2)

Publication Number Publication Date
CN110850420A CN110850420A (en) 2020-02-28
CN110850420B true CN110850420B (en) 2022-05-17

Family

ID=69601527

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911121646.4A Active CN110850420B (en) 2019-11-15 2019-11-15 Fisher SVM sonar signal discrimination method based on marble loss

Country Status (1)

Country Link
CN (1) CN110850420B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111950609A (en) * 2020-07-29 2020-11-17 苏州大学 Sonar signal discrimination system and method based on sparse and robust FisherSVM
CN113030986B (en) * 2021-03-09 2021-09-28 中国科学院水生生物研究所 Method and system for determining isolation degree between different whale populations
CN113900148B (en) * 2021-09-15 2023-09-12 上海微波技术研究所（中国电子科技集团公司第五十研究所） Multi-frequency metal detector and metal target identification method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216879A (en) * 2007-12-28 2008-07-09 西安电子科技大学 Face identification method based on Fisher-supported vector machine
CN103295032B (en) * 2013-05-15 2016-04-13 西安电子科技大学 Based on the image classification method of spatial Fisher vector
CN108805054A (en) * 2018-05-29 2018-11-13 苏州大学 A kind of facial image sorting technique, system, equipment and computer storage media

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7945101B2 (en) * 2007-07-26 2011-05-17 Palo Alto Research Center Incorporated Innovative OCR systems and methods that combine a template based generative model with a discriminative model

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216879A (en) * 2007-12-28 2008-07-09 西安电子科技大学 Face identification method based on Fisher-supported vector machine
CN103295032B (en) * 2013-05-15 2016-04-13 西安电子科技大学 Based on the image classification method of spatial Fisher vector
CN108805054A (en) * 2018-05-29 2018-11-13 苏州大学 A kind of facial image sorting technique, system, equipment and computer storage media

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Support Vector Machine Classifier with Pinball Loss;Xiaolin Huang等;《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》;20140531;第36卷(第5期);正文第1-3页 *

Also Published As

Publication number Publication date
CN110850420A (en) 2020-02-28

Similar Documents

Publication Publication Date Title
Qiao et al. Underwater targets classification using local wavelet acoustic pattern and Multi-Layer Perceptron neural network optimized by modified Whale Optimization Algorithm
CN110850420B (en) Fisher SVM sonar signal discrimination method based on marble loss
Ji et al. Seabed sediment classification using multibeam backscatter data based on the selecting optimal random forest model
Thode et al. Automated detection and localization of bowhead whale sounds in the presence of seismic airgun surveys
CN111965632B (en) Radar target detection method based on Riemann manifold dimensionality reduction
CN105974400A (en) Robust space-time detection method based on symmetric spectral characteristic
Kershenbaum et al. An image processing based paradigm for the extraction of tonal sounds in cetacean communications
CN104732970A (en) Ship radiation noise recognition method based on comprehensive features
US20170294185A1 (en) Segmentation using prior distributions
LeFeuvre et al. Acoustic species identification in the Northwest Atlantic using digital image processing
CN113591733B (en) Underwater acoustic communication modulation mode classification identification method based on integrated neural network model
CN110082822B (en) Method for detecting earthquake by using convolution neural network
CN114117912A (en) Sea clutter modeling and inhibiting method under data model dual drive
CN115114949A (en) Intelligent ship target identification method and system based on underwater acoustic signals
Bore et al. Statistical modelling of extreme ocean current velocity profiles
Hongwei et al. Phase trajectory entropy: A promising tool for passive diver detection
CN115047448A (en) Indoor target rapid detection method and system based on acoustic-electromagnetic intermodulation
Barngrover Automated detection of mine-like objects in side scan sonar imagery
CN114492540A (en) Training method and device of target detection model, computer equipment and storage medium
CN110806444B (en) Seabed sediment recognition and classification method based on shallow stratum profiler and SVM
Cai et al. A reassigned bilinear transformation and Gaussian kernel function-based approach for detecting weak targets in sea clutter
CN111950609A (en) Sonar signal discrimination system and method based on sparse and robust FisherSVM
Kheili et al. Characterization of tropical cyclones to identify the response based design metocean conditions for an FPSO mooring system
KR20220048300A (en) Apparatus and method for generating underwater image data
Hu et al. Recognition of cobalt-rich crusts based on multi-classifier fusion in seafloor mining environments

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant