CN106778809A - A kind of blind source separation method based on improvement chicken group's algorithm - Google Patents

A kind of blind source separation method based on improvement chicken group's algorithm Download PDF

Info

Publication number
CN106778809A
CN106778809A CN201611038289.1A CN201611038289A CN106778809A CN 106778809 A CN106778809 A CN 106778809A CN 201611038289 A CN201611038289 A CN 201611038289A CN 106778809 A CN106778809 A CN 106778809A
Authority
CN
China
Prior art keywords
particle
cock
chicken
signal
hen
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201611038289.1A
Other languages
Chinese (zh)
Inventor
王康
李振璧
姜媛媛
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.)
Anhui University of Science and Technology
Original Assignee
Anhui University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Anhui University of Science and Technology filed Critical Anhui University of Science and Technology
Priority to CN201611038289.1A priority Critical patent/CN106778809A/en
Publication of CN106778809A publication Critical patent/CN106778809A/en
Pending legal-status Critical Current

Links

Classifications

    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Complex Calculations (AREA)

Abstract

The invention discloses a kind of based on the blind source separation method for improving chicken group's algorithm, source signal is obtained into observation signal by nonsingular matrix A first, to obtaining preprocessed signal after observation signal centralization, albefaction, then separation matrix is randomly generated as improvement chicken group's algorithm primary, according to the preprocessed signal for obtaining, particle optimal solution i.e. optimal separation matrix W is obtained using chicken group's algorithm iteration renewal is improved, observation signal is delivered into W obtains optimal separation signal, completes the separation of mixed signal.The inventive method fast convergence rate, high precision is difficult to be absorbed in local optimum, is with a wide range of applications in fields such as signal transacting, radio communications.

Description

A kind of blind source separation method based on improvement chicken group's algorithm
Technical field
The present invention relates to field of signal processing, more particularly to a kind of blind source separation method based on improvement chicken group's algorithm.
Background technology
Blind source separating (blind source separation, BSS) refers in the prior information such as source signal and transmission channel On the premise of unknown, the statistical iteration characteristic according to signal recovers source signal from the mixed signal for observing.
The key problem of blind source separating is the learning algorithm of separation matrix, and its basic thought is made with the feature of statistical iteration It is the expression of input, object function is reached greatly (or minimum).Conventional blind source separation method such as InforMax algorithms, FastICA algorithms etc. are related to the selection of the On The Choice of nonlinear function, these function models to depend primarily on source signal mostly Probability density property, but in practical application the probability density property of source signal be before Signal separator it is unknown, it is non-linear The selection of function greatly affected the separating power of algorithm.
The content of the invention
To solve the above problems, the present invention provides a kind of based on the blind source separation method for improving chicken group's algorithm, and signal is high and steep Negative value after degree absolute value and negentropy weighted average, by improving chicken group's algorithmic minimizing object function, is obtained as object function It is optimal separation matrix to optimal solution, so as to realize the separation to mixing observation signal.The method is to source signal probability density Matter is not relied on, it is adaptable to the separation of all kinds mixed signal, fast convergence rate, high precision, is difficult to be absorbed in local optimum.
To reach above-mentioned purpose, the present invention provides following technical scheme：
A kind of blind source separation method based on improvement chicken group's algorithm, comprises the following steps (1)~(3)：
(1) collection source signal S (t)=[s1(t),s2(t),...,sM(t)]T, nonsingular hybrid matrix A is randomly generated, mix Conjunction matrix A carries out linear hybrid and obtains observation signal X (t)=[x to source signal S (t)1(t),x2(t),...,xM(t)]T, X (t) =AS (t)；Wherein sMT () is the m-th component of source signal S (t), xMT () is the m-th component of observation signal X (t), when t is Between sequence, subscript T represents conjugate transposition, and M is positive integer, and A is M × M dimension matrixes；
(2) preprocessed signal Z (t) is obtained after carrying out centralization and albefaction to observation signal X (t) obtained in step (1), Centralization and albefaction are existing mature technology, and here is omitted for concrete principle；
(3) separation matrix is randomly generated as chicken group's algorithm primary is improved, according to preprocessed signal Z (t) for obtaining, Optimal solution i.e. optimal separation matrix W is obtained using chicken group's algorithm iteration renewal is improved, X (t) is delivered into W obtains optimal separation signal Y (t)=[y1(t),y2(t),...,yM(t)]T, Y (t)=WX (t), the separation of completion mixed signal.
A kind of utilization based on the blind source separation method for improving chicken group's algorithm, in the step (3) of the invention improves chicken group Algorithm obtains comprising the following steps that for the i.e. optimal separation matrix W of optimal solution：
(3.1) initialization chicken group, sets maximum iteration t1=M1, chicken group population N=100, randomly generate separation square Battle array defines cock particle number N as chicken group's particleR=0.15N, hen particle number NH=0.7N, chicken particle number NC= 0.25N, mother's hen particle number NM=0.5NH；
(3.2) fitness function fitness is set, pretreated signal Z (t) is delivered to the separation matrix for randomly generating (chicken group particle) obtains initially-separate signal, and centralization, whitening operation are carried out to initially-separate signal, brings fitness function into Fitness calculates the fitness value of chicken group's particle, the setting current desired positions of particle and the global desired positions of chicken group, chicken group's algorithm Iterations t1=1；
(3.3) if t1/ G=1 (i.e. currently be the first generation), sort fitness value and according to cock, hen from small to large With the division that chicken particle number determines cock, hen and chicken, chicken group's hierarchy is set up, chicken group is divided into several subgroups simultaneously Determine hen particle and chicken particle corresponding mother-child relationship (MCR) (if having in each subgroup a cock particle and dried hen particle and Chicken particle is constituted), wherein, G represents the algebraically for starting to update hierarchy, dominance relation and mother-child relationship (MCR), G=10；
(3.4) according to formula (1)：
xir,j(t1+ 1)=xir,j(t1)·(1+Φ(0,σ2))………………………………(1)
To update cock particle position, wherein, xir,j(t1),xir,j(t1+ 1) cock particle ir is represented respectively in t1It is secondary and t1In the location of jth dimension space in+1 iteration；Φ(0,σ2) it is that a variance is σ2Gaussian Profile, σ2Expression formula is：
Wherein, firAnd fkrThe fitness value of cock particle ir and cock particle kr is represented, ε is a minimum constant, used To ensure that denominator is meaningful, NRIt is whole chicken group's cock number of particles, kr is that any one after ir is removed in all cock particles Body, when the fitness value of cock particle ir is better than the fitness value of cock particle kr, variances sigma2It is 1, the search of cock particle ir Space becomes big, on the contrary σ2Reduce, the search space of cock particle ir reduces；
(3.5) hen particle will follow the cock particle of its subgroup to scan for, while also following the cock of other subgroups Particle is scanned for, according to formula (3)
xih,j(t1+ 1)=xih,j(t1)+C1·θ·(xr1,j(t1)-xih,j(t1))+C2·θ·(xr2,j(t1)-xih,j (t1))……(3)
Hen particle position is updated, wherein, xr1,j(t1),xr2,j(t1) cock of the affiliated subgroup of hen particle is represented respectively The positional information of particle and other subgroup cock particles, θ is equally distributed random number, C between 0 to 11And C2Mother is represented respectively Chicken particle refers to itself subgroup and other subgroup weights, according to formula (4), (5)
C1=exp ((fih-fr1)/(abs(fir+ε))…………………………………(4)
C2=exp ((fr2-fir))……………………………………(5)
Obtain, wherein, fihAnd fr1Hen particle ih and affiliated subgroup cock particle r is represented respectively1Fitness value, fr2 The fitness value of other subgroup cock particles that representative is randomly selected；
(3.6) chicken particle not only follows mother's hen particle of its subgroup to scan for, while to place subgroup cock Study, according to formula (6)
xic,j(t1+ 1)=wxic,j(t1)+F·(xm,j(t1)-xic,j(t1))+C·(xr,j(t1)-xic,j (t1))…………(6)
Chicken position is updated, wherein, xm,j(t1) represent the positional information that chicken particle follows mother's hen particle, xr,j (t1) cock particle location information in subgroup where mother hen particle itself is represented, C is Studying factors, and value 0.5 is represented The degree that chicken particle learns to cock particle in subgroup where itself, w is chicken particle self inertia weight, here using certainly Inertia weight is adapted to, by formula (7)
W=(wmax-wmin) * exp (- (τ * ((t1-1)/(M1-1))2)+wmin……………………(7)
Obtain particle itself adaptive weighting, wherein, wmax is the maximum of inertia weight, wmin be inertia weight most Small value, τ takes 50, F to follow coefficient, represents that chicken particle follows mother's hen particle search of food；
(3.7) using the adaptation that each particle is calculated after formula (1)~(7) renewal cock, hen and chicken particle position Angle value, updates the current desired positions of particle and the global desired positions of chicken group of chicken group；
(3.8)t1=t1+ 1, if reaching iterations, stop iteration, obtain optimal location (optimal solution), i.e., optimal point From matrix W, step (3.3) is otherwise gone to.
It is of the invention a kind of based on the blind source separation method for improving chicken group's algorithm, improvement chicken group's algorithm in the step (3) Fitness function fitness is the weighted average for separating signal negentropy and kurtosis absolute value, is implemented as：
Fitness=- (| 0.4 × fitness1+0.6 × fitness2 |/2) ... ... ... ... (8)
In formula (8),Signal negentropy is represented, whereinYi represents i-th Individual separation signal；Represent the absolute value of signal kurtosis, wherein kurt (yi) it is i-th kurtosis of separation signal, fitness function value is smaller to show that separating effect is better.
Beneficial effects of the present invention：The weighted average of signal negentropy and kurtosis absolute value as traditional chicken group's algorithm will be separated Object function, it is to avoid the conventional blind source separation method nonlinear function that for example InforMax algorithms, FastICA algorithms etc. are related to On The Choice, meanwhile, to traditional chicken group's algorithm improvement, chicken particle position is there is self adaptation inertia weight and to place subgroup The ability of cock study, overcomes traditional chicken group algorithm chicken particle to be easily absorbed in local defect, and the inventive method has convergence speed Degree is fast, the small advantage of crosstalk error, and in radio communication, the field such as signal transacting is with a wide range of applications.
Brief description of the drawings
Fig. 1 is a kind of based on the blind source separation method overall flow figure for improving chicken group's algorithm
Fig. 2 is to improve chicken group's algorithm flow chart
Specific embodiment
The present invention provides a kind of based on the blind source separation method for improving chicken group's algorithm, and the method is to source signal probability density Matter is not relied on, it is adaptable to the separation of all kinds mixed signal, fast convergence rate, high precision, is difficult to be absorbed in local optimum.
To reach above-mentioned purpose, specific embodiment of the invention is as follows：
(1) collection source signal S (t)=[s1(t),s2(t),...,sM(t)]T, nonsingular hybrid matrix A is randomly generated, mix Conjunction matrix A carries out linear hybrid and obtains observation signal X (t)=[x to source signal S (t)1(t),x2(t),...,xM(t)]T,X(t) =AS (t)；Wherein sMT () is the m-th component of source signal S (t), xMT () is the m-th component of observation signal X (t), when t is Between sequence, subscript T represents conjugate transposition, and M is positive integer, and A is M × M dimension matrixes；
(2) preprocessed signal Z (t) is obtained after carrying out centralization and albefaction to observation signal X (t) obtained in step (1), Centralization and albefaction are existing mature technology, and here is omitted for concrete principle；
(3) separation matrix is randomly generated as chicken group's algorithm primary is improved, according to preprocessed signal Z (t) for obtaining, Optimal solution i.e. optimal separation matrix W is obtained using chicken group's algorithm iteration renewal is improved, X (t) is delivered into W obtains optimal separation signal Y (t)=[y1(t),y2(t),...,yM(t)]T, Y (t)=WX (t), the separation of completion mixed signal.Chicken group's algorithm is improved to obtain Optimal solution is comprising the following steps that for optimal separation matrix W：
(3.1) initialization chicken group, sets maximum iteration t1=M1, chicken group population N=100, randomly generate separation square Battle array defines cock particle number N as chicken group's particleR=0.15N, hen particle number NH=0.7N, chicken particle number NC= 0.25N, mother's hen particle number NM=0.5NH；
(3.2) fitness function fitness is set, and fitness function fitness is absolute with kurtosis to separate signal negentropy The weighted average of value, is implemented as：Fitness=- (| 0.4 × fitness1+0.6 × fitness2 |/2), whereinSignal negentropy is represented, whereinyiRepresent i-th separation signal；Represent the absolute value of signal kurtosis, wherein kurt (yi) it is i-th The kurtosis of signal is separated, fitness function value is smaller to show that separating effect is better.Pretreated signal Z (t) is delivered at random The separation matrix (chicken group particle) of generation obtains initially-separate signal, and centralization, whitening operation, band are carried out to initially-separate signal Enter the fitness value that fitness function fitness calculates chicken group's particle, the setting current desired positions of particle and chicken group are global best Position, chicken group's algorithm iteration number of times t1=1；
(3.3) if t1/ G=1 (i.e. currently be the first generation), sort fitness value and according to cock, hen from small to large With the division that chicken particle number determines cock, hen and chicken, chicken group's hierarchy is set up, chicken group is divided into several subgroups simultaneously Determine hen particle and chicken particle corresponding mother-child relationship (MCR) (if having in each subgroup a cock particle and dried hen particle and Chicken particle is constituted), wherein, G represents the algebraically for starting to update hierarchy, dominance relation and mother-child relationship (MCR), G=10；
(3.4) according to formula (1)：
xir,j(t1+ 1)=xir,j(t1)·(1+Φ(0,σ2))………………………………(1)
To update cock particle position, wherein, xir,j(t1),xir,j(t1+ 1) cock particle ir is represented respectively in t1It is secondary and t1In the location of jth dimension space in+1 iteration；Φ(0,σ2) it is that a variance is σ2Gaussian Profile, σ2Expression formula is：
Wherein, firAnd fkrThe fitness value of cock particle ir and cock particle kr is represented, ε is a minimum constant, used To ensure that denominator is meaningful, NRIt is whole chicken group's cock number of particles, kr is that any one after ir is removed in all cock particles Body, when the fitness value of cock particle ir is better than the fitness value of cock particle kr, variances sigma2It is 1, the search of cock particle ir Space becomes big, on the contrary σ2Reduce, the search space of cock particle ir reduces；
(3.5) hen particle will follow the cock particle of its subgroup to scan for, while also following the cock of other subgroups Particle is scanned for, according to formula (3)
xih,j(t1+ 1)=xih,j(t1)+C1·θ·(xr1,j(t1)-xih,j(t1))+C2·θ·(xr2,j(t1)-xih,j (t1))……(3)
Hen particle position is updated, wherein, xr1,j(t1),xr2,j(t1) cock of the affiliated subgroup of hen particle is represented respectively The positional information of particle and other subgroup cock particles, θ is equally distributed random number, C between 0 to 11And C2Mother is represented respectively Chicken particle refers to itself subgroup and other subgroup weights, according to formula (4), (5)
C1=exp ((fih-fr1)/(abs(fir+ε))……………………………(4)
C2=exp ((fr2-fir))…………………………………(5)
Obtain, wherein, fihAnd fr1Hen particle ih and affiliated subgroup cock particle r is represented respectively1Fitness value, fr2 The fitness value of other subgroup cock particles that representative is randomly selected；
(3.6) chicken particle not only follows mother's hen particle of its subgroup to scan for, while to place subgroup cock Study, according to formula (6)
xic,j(t1+ 1)=wxic,j(t1)+F·(xm,j(t1)-xic,j(t1))+C·(xr,j(t1)-xic,j (t1))……………(6)
Chicken position is updated, wherein, xm,j(t1) represent the positional information that chicken particle follows mother's hen particle, xr,j (t1) cock particle location information in subgroup where mother hen particle itself is represented, C is Studying factors, and value 0.5 is represented The degree that chicken particle learns to cock particle in subgroup where itself, w is chicken particle self inertia weight, here using certainly Inertia weight is adapted to, by formula (7)
W=(wmax-wmin) * exp (- (τ * ((t1-1)/(M1-1))2)+wmin……………………(7)
Obtain particle itself adaptive weighting, wherein, wmax is the maximum of inertia weight, wmin be inertia weight most Small value, τ takes 50, F to follow coefficient, represents that chicken particle follows mother's hen particle search of food；
(3.7) using the adaptation that each particle is calculated after formula (1)~(7) renewal cock, hen and chicken particle position Angle value, updates the current desired positions of particle and the global desired positions of chicken group of chicken group；
(3.8)t1=t1+ 1, if reaching iterations, stop iteration, obtain optimal location (optimal solution), i.e., optimal point From matrix W, step (3.3) is otherwise gone to.
Above example is only explanation technological thought of the invention, it is impossible to once limit protection scope of the present invention, every According to technological thought proposed by the present invention, any change done on the basis of technical method each falls within the scope of the present invention Within.

Claims (3)

1. it is a kind of based on the blind source separation method for improving chicken group's algorithm, it is characterised in that to comprise the following steps：
(1) collection source signal S (t)=[s1(t),s2(t),...,sM(t)]T, randomly generate nonsingular hybrid matrix A, mixed moment Battle array A carries out linear hybrid and obtains observation signal X (t)=[x to source signal S (t)1(t),x2(t),...,xM(t)]T, X (t)=AS (t)；Wherein sMT () is the m-th component of source signal S (t), xMT () is the m-th component of observation signal X (t), t is time sequence Row, subscript T represents conjugate transposition, and M is positive integer, and A is M × M dimension matrixes；
(2) preprocessed signal Z (t), center are obtained after carrying out centralization and albefaction to observation signal X (t) obtained in step (1) Change and albefaction is existing mature technology, here is omitted for concrete principle；
(3) separation matrix is randomly generated as chicken group's algorithm primary is improved, and according to preprocessed signal Z (t) for obtaining, is utilized Improve chicken group's algorithm iteration renewal and obtain optimal solution i.e. optimal separation matrix W, X (t) is delivered into W obtains optimal separation signal Y (t) =[y1(t),y2(t),...,yM(t)]T, Y (t)=WX (t), the separation of completion mixed signal.
2. it is as claimed in claim 1 a kind of based on the blind source separation method for improving chicken group's algorithm, it is characterised in that the step (3) utilization improves chicken group's algorithm and obtains comprising the following steps that for the i.e. optimal separation matrix W of optimal solution in：
(3.1) initialization chicken group, sets maximum iteration t1=M1, chicken group population N=100, randomly generate separation matrix work It is chicken group's particle, defines cock particle number NR=0.15N, hen particle number NH=0.7N, chicken particle number NC= 0.25N, mother's hen particle number NM=0.5NH；
(3.2) fitness function fitness is set, pretreated signal Z (t) is delivered to the separation matrix (chicken for randomly generating Group's particle) initially-separate signal is obtained, centralization, whitening operation are carried out to initially-separate signal, substitute into fitness function Fitness calculates the fitness value of chicken group's particle, the setting current desired positions of particle and the global desired positions of chicken group, chicken group's algorithm Iterations t1=1；
(3.3) if t1/ G=1 (i.e. currently be the first generation), sort fitness value and according to cock, hen and chicken from small to large Particle number determines the division of cock, hen and chicken, sets up chicken group's hierarchy, is divided into several subgroups and determines mother chicken group The corresponding mother-child relationship (MCR) of chicken particle and chicken particle is (if having a cock particle and dried hen particle and chicken grain in each subgroup Son is constituted), wherein, G represents the algebraically for starting to update hierarchy, dominance relation and mother-child relationship (MCR), G=10；
(3.4) according to formula (1)：
xir,j(t1+ 1)=xir,j(t1)·(1+Φ(0,σ2))……………………………(1)
To update cock particle position, wherein, xir,j(t1),xir,j(t1+ 1) cock particle ir is represented respectively in t1Secondary and t1+1 In the location of jth dimension space in secondary iteration；Φ(0,σ2) it is that a variance is σ2Gaussian Profile, σ2Expression formula is：
σ 2 = 1 i f f i r ≤ f k r exp ( f k r - f i r | f i r | + ϵ ) e l s e , k r ∈ [ 1 , N R ] , k r ≠ i r ... ( 2 )
Wherein, firAnd fkrThe fitness value of cock particle ir and cock particle kr is represented, ε is a minimum constant, for protecting Card denominator is meaningful, NRIt is whole chicken group's cock number of particles, kr is that any individuality after ir is removed in all cock particles, when The fitness value of cock particle ir is better than the fitness value of cock particle kr, variances sigma2It is 1, the search space of cock particle ir becomes Greatly, on the contrary σ2Reduce, the search space of cock particle ir reduces；
(3.5) hen particle will follow the cock particle of its subgroup to scan for, while also following the cock particle of other subgroups Scan for, according to formula (3)
xih,j(t1+ 1)=xih,j(t1)+C1·θ·(xr1,j(t1)-xih,j(t1))+C2·θ·(xr2,j(t1)-xih,j(t1))…… (3)
Hen particle position is updated, wherein, xr1,j(t1),xr2,j(t1) represent respectively the affiliated subgroup of hen particle cock particle and The positional information of other subgroup cock particles, θ is equally distributed random number, C between 0 to 11And C2Hen particle is represented respectively With reference to itself subgroup and other subgroup weights, according to formula (4), (5)
C1=exp ((fih-fr1)/(abs(fir+ε))…………………………(4)
C2=exp ((fr2-fir))…………………………(5)
Obtain, wherein, fihAnd fr1Hen particle ih and affiliated subgroup cock particle r is represented respectively1Fitness value, fr2Represent with The fitness value of other subgroup cock particles that machine is chosen；
(3.6) chicken particle not only follows mother's hen particle of its subgroup to scan for, while learn to place subgroup cock, According to formula (6)
xic,j(t1+ 1)=wxic,j(t1)+F·(xm,j(t1)-xic,j(t1))+C·(xr,j(t1)-xic,j(t1))………… (6)
Chicken position is updated, wherein, xm,j(t1) represent the positional information that chicken particle follows mother's hen particle, xr,j(t1) The cock particle location information in subgroup where mother hen particle itself is represented, C is Studying factors, and value 0.5 represents chicken The degree that particle learns to cock particle in subgroup where itself, w is chicken particle self inertia weight, here using self adaptation Inertia weight, by formula (7)
Particle itself adaptive weighting is obtained, wherein, wmax is the maximum of inertia weight, and wmin is the minimum of inertia weight Value, τ takes 50, F to follow coefficient, represents that chicken particle follows mother's hen particle search of food；
(3.7) using the fitness value that each particle is calculated after formula (1)~(7) renewal cock, hen and chicken particle position, Update the current desired positions of particle and the global desired positions of chicken group of chicken group；
(3.8)t1=t1+ 1, if reaching iterations, stop iteration, obtain optimal location (optimal solution), i.e. optimal separation square Battle array W, otherwise goes to step (3.3).
3. it is as claimed in claim 2 a kind of based on the blind source separation method for improving chicken group's algorithm, it is characterised in that the improvement Chicken group algorithm fitness function fitness is the weighted average for separating signal negentropy and kurtosis absolute value, is implemented as：
Fitness=- (0.4 × fitness1+0.6 × fitness2 |/2) ... ... ... ... (8)
In formula (8),Signal negentropy is represented, whereinyiRepresent i-th point From signal；Represent the absolute value of signal kurtosis, wherein kurt (yi) It is i-th kurtosis of separation signal, fitness function value is smaller to show that separating effect is better.
CN201611038289.1A 2016-11-23 2016-11-23 A kind of blind source separation method based on improvement chicken group's algorithm Pending CN106778809A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611038289.1A CN106778809A (en) 2016-11-23 2016-11-23 A kind of blind source separation method based on improvement chicken group's algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611038289.1A CN106778809A (en) 2016-11-23 2016-11-23 A kind of blind source separation method based on improvement chicken group's algorithm

Publications (1)

Publication Number Publication Date
CN106778809A true CN106778809A (en) 2017-05-31

Family

ID=58974934

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611038289.1A Pending CN106778809A (en) 2016-11-23 2016-11-23 A kind of blind source separation method based on improvement chicken group's algorithm

Country Status (1)

Country Link
CN (1) CN106778809A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107153800A (en) * 2017-05-04 2017-09-12 天津工业大学 A kind of reader antenna Optimization deployment scheme that alignment system is recognized based on the super high frequency radio frequency for improving chicken group's algorithm
CN107358619A (en) * 2017-09-07 2017-11-17 陕西师范大学 Multi-Level Threshold Image Segmentation method based on chicken group's optimization
CN107682117A (en) * 2017-09-11 2018-02-09 天津工业大学 A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm
CN107741410A (en) * 2017-09-18 2018-02-27 燕山大学 A kind of fluorescence spectrum detection method of the benzo class material based on ICSO SVM
CN110119778A (en) * 2019-05-10 2019-08-13 辽宁大学 A kind of equipment method for detecting health status improving chicken group's optimization RBF neural
CN110208375A (en) * 2019-06-13 2019-09-06 石家庄铁道大学 A kind of detection method and terminal device of anchor rod anchored defect
WO2020037464A1 (en) * 2018-08-20 2020-02-27 唐山照澜海洋科技有限公司 Gene regulatory network construction method based on ensemble feature importance and chicken swarm algorithm
CN114936577A (en) * 2022-05-23 2022-08-23 大连大学 Mixed image blind separation method based on improved lion group algorithm

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107153800A (en) * 2017-05-04 2017-09-12 天津工业大学 A kind of reader antenna Optimization deployment scheme that alignment system is recognized based on the super high frequency radio frequency for improving chicken group's algorithm
CN107358619A (en) * 2017-09-07 2017-11-17 陕西师范大学 Multi-Level Threshold Image Segmentation method based on chicken group's optimization
CN107682117A (en) * 2017-09-11 2018-02-09 天津工业大学 A kind of design method based on the long LT codes degree distribution of short code for improving chicken colony optimization algorithm
CN107741410A (en) * 2017-09-18 2018-02-27 燕山大学 A kind of fluorescence spectrum detection method of the benzo class material based on ICSO SVM
WO2020037464A1 (en) * 2018-08-20 2020-02-27 唐山照澜海洋科技有限公司 Gene regulatory network construction method based on ensemble feature importance and chicken swarm algorithm
CN110119778A (en) * 2019-05-10 2019-08-13 辽宁大学 A kind of equipment method for detecting health status improving chicken group's optimization RBF neural
CN110119778B (en) * 2019-05-10 2024-01-05 辽宁大学 Equipment health state detection method for improving chicken flock optimization RBF neural network
CN110208375A (en) * 2019-06-13 2019-09-06 石家庄铁道大学 A kind of detection method and terminal device of anchor rod anchored defect
CN114936577A (en) * 2022-05-23 2022-08-23 大连大学 Mixed image blind separation method based on improved lion group algorithm
CN114936577B (en) * 2022-05-23 2024-03-26 大连大学 Mixed image blind separation method based on improved lion group algorithm

Similar Documents

Publication Publication Date Title
CN106778809A (en) A kind of blind source separation method based on improvement chicken group's algorithm
CN108564006B (en) Polarized SAR terrain classification method based on self-learning convolutional neural network
CN108038445B (en) SAR automatic target identification method based on multi-view deep learning framework
Jin et al. Data normalization to accelerate training for linear neural net to predict tropical cyclone tracks
CN102651073B (en) Sparse dynamic ensemble selection-based SAR (synthetic aperture radar) image terrain classification method
CN104166859B (en) Based on SSAE and FSALS SVM Classification of Polarimetric SAR Image
CN104750982A (en) Reliability interval estimation method based on proportion grouping re-sampling
CN111582236A (en) LPI radar signal classification method based on dense convolutional neural network
CN101547172B (en) RBF-based blind equalization method for fuzzy-neural network
CN106202756A (en) Based on monolayer perceptron owing determines blind source separating source signal restoration methods
CN114019467A (en) Radar signal identification and positioning method based on MobileNet model transfer learning
CN103440275A (en) Prim-based K-means clustering method
CN113469198A (en) Image classification method based on improved VGG convolutional neural network model
Zhang et al. Weighted random forest algorithm based on bayesian algorithm
CN108388961A (en) Self-adapting random neighbours' community detecting algorithm based on modularity optimization
CN110954862B (en) Radiation source direct positioning method based on global narrow-band model under sparse Bayesian framework
CN106534009A (en) Improved variable step size equivariant adaptive blind source separation method
CN110596668B (en) Target external radiation source passive positioning method based on reciprocal deep neural network
CN104881688A (en) Two-stage clustering algorithm based on difference evolution and fuzzy C-means
CN115294381B (en) Small sample image classification method and device based on feature migration and orthogonal prior
Yang et al. Short-term wind power prediction based on CEEMDAN and parallel CNN-LSTM
CN107563334A (en) Based on the face identification method for differentiating linear expression retaining projection
Panda et al. A WNN model trained with orthogonal colliding bodies optimization for accurate identification of hammerstein plant
CN106997486A (en) A kind of higher-dimension multi-objective optimization algorithm split based on space
You et al. Prediction of Tropical Pacific Rain Rates with Over-parameterized Neural Networks

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20170531