CN105404783A - Blind source separation method - Google Patents
Blind source separation method Download PDFInfo
- Publication number
- CN105404783A CN105404783A CN201510861672.6A CN201510861672A CN105404783A CN 105404783 A CN105404783 A CN 105404783A CN 201510861672 A CN201510861672 A CN 201510861672A CN 105404783 A CN105404783 A CN 105404783A
- Authority
- CN
- China
- Prior art keywords
- particle
- fitness
- matrix
- formula
- blind source
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Feedback Control In General (AREA)
Abstract
The present invention provides a blind source separation method, and belongs to the technical field of signal processing. The method comprises three steps of: blind source problem problem modeling, fitness evaluation, and separation matrix solving. According to the blind source separation method provided by the present invention, a parameter-adaptive particle swarm algorithm is adopted, and chaotic iteration and a cloud model are introduced into the particle swarm algorithm, so that a particle swarm alternate between chaos and stability and gets close to an optimal solution, thereby effectively solving problems that solving a separation matrix in a blind source separation problem is prone to fall into a local optimum solution and a premature convergence problem, greatly reducing a search time, and reducing time complexity of the blind source separation.
Description
Technical field
The invention belongs to signal transacting field, particularly a kind of blind source separation method of the parameter adaptive particle cluster algorithm based on cloud model and chaos.
Background technology
Blind source separating (BlindSourceSeperation, BSS) be a kind of new signal processing method that last century, the eighties grew up along with rising once again of neural network, its thought comes from the research of people to " cocktail party ", its essence is when source signal and transport channel parameters are all unknown, only according to the statistical property of input signal, recovered the process of source signal by observation signal.
independentcomponent analysis (IndependentComponentAnalysis, ICA) as the main method of blind source separating, comprise objective function choose and optimize two parts, traditional IC A optimizes employing steepest gradient descent algorithm, there is speed of convergence and be absorbed in the difficult problems such as locally optimal solution slowly, easily, the quality of separating cannot be guaranteed, and cause in actual applications, effect of signal separation is poor.
Particle cluster algorithm (ParticleSwarmOptimization, PSO) is the main optimization method of current blind source separating, is to be obtained by the inspiration of flock of birds foraging behavior, is a kind of optimized algorithm based on swarm intelligence.The advantages such as basic particle cluster algorithm has that parameter is simple, fast convergence rate, hunting zone are large, but because this algorithm carries out initialization operation at random, the quality of particle can not be guaranteed, the optimum solution being easy to a part of particle distance is sought is far, causes speed of searching optimization greatly to reduce.Moreover self principle of particle cluster algorithm also causes search abundant not, and what obtain is not optimum solution, and can not ensure that when solution space is larger each position is all searched especially, particle is easily absorbed in locally optimal solution, is unfavorable for the optimization efficiency improving algorithm.
Summary of the invention
Technical matters to be solved by this invention is: be easily absorbed in locally optimal solution and precocious problem for blind source separate technology, propose a kind of parameter adaptive particle cluster algorithm based on cloud model and chaos and ask optimum separation matrix, and then realize the method for blind source separating, to improve separation accuracy.
The technology of the present invention problem can be achieved through the following technical solutions:
A kind of blind source separation method, has the modeling of blind source separating problem, fitness evaluation and separation matrix to solve three steps:
Described blind source separating problem modeling is in statistics
independentassumed condition under, recovered the process of source signal by observation signal; Particular by searching separation matrix, each vector of output matrix obtained after making this matrix and observing the signal matrix obtained be multiplied is mutual
independent, then this output matrix is exactly the estimated signal of source signal;
Described fitness evaluation is, after particle initialization, the fitness according to particle is divided into two populations particle, according to formula kurt (the y)=E{y asking fitness
4-3 (E{y
2)
2, obtain the average fitness of all particles, the particle being greater than average fitness adopts Chaos particle swarm optimization algorithm to try to achieve global optimum position; The particle being less than average fitness adopts cloud model particle cluster algorithm to generate the method for inertia weight, tries to achieve global optimum position; Described particle particle cluster algorithm initialization procedure on emulation platform produces at random; Described initialization refers to, first the matrix of stochastic generation n × n dimension is as initially-separate matrix, using the initial global optimum position of each for this matrix column vector as particle, limit in interval in Studying factors, inertia weight and speed, all particles are drawn close to the personal best particle making fitness maximum and are upgraded personal best particle, all personal best particles are arranged as the matrix of a global optimum position, this matrix is the separation matrix of n × n dimension;
Described separation matrix solves and is, fitness is greater than to the particle of average fitness, Chaos particle swarm optimization algorithm is adopted to solve, particularly utilize chaology generate one group with the same number of Chaos Variable of problem control variable to be asked, chaotic disturbance is carried out to control variable, the traversal scope of chaos is transformed into the restriction range of control variable, finally, according to speed and the location updating formula iteration of population, seek the globally optimal solution of problem; Particle fitness being less than to average fitness adopts cloud model particle cluster algorithm to solve, X condition cloud generator is particularly utilized to generate water dust as particle, and adaptive generation inertia weight, the speed of recycling particle cluster algorithm and location updating formula, upgrade the individual extreme value of water dust and global optimum, thus find the position making fitness reach maximum one group of water dust to separate as output; The optimal location that comparison two kinds of optimizing algorithms are obtained obtains global optimum position, and this optimal location is also globally optimal solution, solves the separation matrix of mixed matrix as blind source separating.
The mathematical modeling of blind source separating problem of the present invention, particularly according to formula X (t)=AS (t)+n (t) and
modeling is carried out, wherein X (t)=[x to the instantaneous mixing of blind source separating and Signal separator
1, x
2..., x
m]
tfor observation signal set, x
i∈ C
mfor each road signal of actual observation, S (t)=[s
1, s
2..., s
n]
t, s
i∈ C
nfor mutual statistical
independentsource signal, n (t) be m × n tie up additive noise, A is hybrid matrix, and W is separation matrix.
Employing Chaos particle swarm optimization algorithm of the present invention solves, and is introduce chaos iteration in particle cluster algorithm, particularly according to logical mappings formula Z
n+1=4Z
n(1-Z
n) represent, in formula, Z
nrepresent Chaos Variable, according to chaos principle, chaotic disturbance is added to population, i.e. Z'
k=(1-β) Z
*+ β Z
k, in formula, Z
kfor chaos vector when k time, Z'
kfor adding the chaos vector after disturbance, β ∈ [0,1] represents the intensity of disturbance, adopts self-adaptation value, and at the search initial stage, its value is comparatively large, strengthens the disturbance to solution vector, and along with going deep into of search, β slowly reduces, and is specifically changed to
in formula, n is an integer.
Employing cloud model particle cluster algorithm of the present invention solves, and is the inertia weight being generated population by following formula; E
n=f
avg', E
n=(f
avg'-f
g') ÷ c
c1, H
e=E
n÷ c
c2, E
n'=normrnd (E
n, H
e); Wherein f
avg' be the average fitness of this part population adopting cloud model; f
g' be the optimal-adaptive degree of this part population; c
c1, c
c2for the controling parameters of cloud model.
Particle cluster algorithm of the present invention refers to have parameter adaptive particle cluster algorithm, particularly according to formula v
id(t+1)=ω v
id(t)+c
1r
1(p
id-x
id(t))+c
2r
2(p
gd-x
id(t)) and formula x
id(t+1)=x
id(t)+v
id(t+1) the renewal iteration of position and speed is carried out, wherein i=1,2 ..., M, M are the scale of population, namely total number of particles; D=1,2 ..., N, N are the dimensions of search volume; v
idthe speed of particle i when d dimension; x
idthe position of particle i when d dimension; ω is inertia weight; c
1and c
2it is Studying factors; p
idand p
gdlocal optimum position and the global optimum position of particle i respectively; r
1and r
2be the random number be distributed on [0,1], described parameter adaptive, comprise Studying factors c
1, c
2, wherein c
1(t)=2.5-2*exp (-α | f
avg-f
g|), c
2(t)=0.5+2*exp (-α | f
avg-f
g|), f
goptimal-adaptive degree, f
avgbe the average fitness of population, α is control coefrficient; Adaptive generation inertia weight ω,
in formula, iter
maxit is maximum evolution number of times; The fitness variance of each particle is σ
2,
in formula
f
iit is the fitness of i-th particle.
Core concept of the present invention is: stochastic generation population, and in setting range the position of initialization particle and speed, obtain the fitness value of each particle, obtain the individual optimal value of each particle, obtain global optimum by comparing, particle cluster algorithm (PSO) is a kind of optimized algorithm based on swarm intelligence, and simulation flock of birds is looked for food, the peripheral region of the current bird nearest from food is searched, the position of search of food again after reducing the scope by the cooperation of collective between bird; In target search scope, the solution of each optimization problem is called " particle ", each particle position, speed and fitness (fitnessvalue) three indexs characterize, and all particles come renewal speed and position according to the history optimal location of self and the optimal location of whole particle colony; By calculating the target function value of particle, weigh the good and bad degree of particle with this fitness.First obtain the average fitness of all particles, require population to be divided into two classes according to evaluation, be respectively outstanding particle and general particle.The particle of different population adopts different optimizing algorithms: outstanding particle will adopt Chaos particle swarm optimization algorithm, carrying out chaos iteration, finding the globally optimal solution of this population to being absorbed in precocious particle; General particle adopts cloud model particle cluster algorithm, is found the globally optimal solution of this population by cloud model, and cloud model (Cloudmodel) is a kind of
novelqualitative and quantitative between uncertain transformation model, by fuzzy mathematics is combined with probabilistic models, adopt general normal distribution to represent and be difficult to the uncertain description that represents by exact numerical values recited; Cloud model is by three mathematical feature (E
x, E
n, H
e) represent, wherein E
xbe expected value, represent the expectation value of water dust on domain; E
nbe entropy, represent the uncertainty of water dust; H
ebeing super entropy, is the uncertainty measure of entropy, is jointly determined by the ambiguity of entropy and randomness; Cloud model particle cluster algorithm is under the control of three numerical characteristics of given cloud model, according to normal characteristics self-adaptative adjustment inertia weight, hunts out optimum solution.Finally by comparing, find globally optimal solution.This optimum solution is multiplied with mixed signal as separation matrix, obtains isolated each road signal, complete blind source separating.
To sum up, the present invention has following beneficial effect:
What 1, algorithm of the present invention adopted introduces chaos iteration in particle cluster algorithm, make population chaos and stable between close alternately to optimum solution, effectively solve and ask separation matrix to be easily absorbed in locally optimal solution and premature problem in blind source separating problem.
What 2, algorithm of the present invention adopted introduces cloud model in particle cluster algorithm, greatly shortens search time, reduces the time complexity of blind source separating.
What 3, algorithm of the present invention adopted is divided into two classes by particle according to different adaptive values, uses different inertia weights respectively, improves the diversity of population and the ergodicity of particle search while not changing particle swarm optimization algorithm essence.
4, the parameter adaptive particle cluster algorithm of algorithm employing of the present invention, ensure that searching process can be weighed between search search precision and time efficiency and carries out.
Accompanying drawing explanation
fig. 1it is blind source separating model of the present invention.
fig. 2it is overall flow of the present invention signal
figure.
fig. 3it is the idiographic flow of Chaos particle swarm optimization algorithm of the present invention
figure.
fig. 4it is the idiographic flow of chaotic disturbance algorithm of the present invention
figure.
fig. 5it is the idiographic flow of cloud model particle cluster algorithm of the present invention
figure.
Specific embodiments
Below according to instructions
accompanying drawingand specific embodiment, the present invention is further described:
Embodiment 1, mathematical model of the present invention.
Reference
fig. 1, a kind of blind source separation method of the present invention, the mathematical modeling in the present embodiment is: be provided with n source signal, and they form n-dimensional vector: S (t)=[s
1, s
2..., s
n]
t, s
i∈ C
nfor mutual statistical
independentsource signal; Mutual statistical between vector
independent, m mixed signal forms m and ties up observation data vector: X (t)=[x
1, x
2..., x
m]
t, x
i∈ C
mfor the signal of actual observation; The mathematical model of noisy Blind Signal Separation is:
X(t)＝A·S(t)+n(t)(1)
In formula, A is the hybrid matrix of m × n dimension, and n (t) is m × n dimension additive noise.
Blind source separating problem-solving approach is when ignoring noise n (t), finds a separation matrix W, and matrix Y (t) after separation is met:
Here adopt kurtosis as objective function:
kurt(y)＝E{y
4}-3(E{y
2})
2(3)
The non-Gaussian system of signal is stronger, and the absolute value of kurtosis is larger, and the present invention adopts particle swarm optimization algorithm to find and can make | kurt (Y (t)) | and be worth one group of maximum solution as separation matrix W, obtain the estimated signal of original signal with this.
Embodiment 2, overall step of the present invention.
Reference
fig. 2, a kind of blind source separation method of the present invention, in the present embodiment, overall step of the present invention is:
Step one: pre-service is carried out to the signal of observation.
Described pre-service comprises carries out centralization and albefaction to mixed signal, and centralization is also called average, and available following formula realizes:
Carrying out whitening operation to mixed signal, is uncorrelated between signal, the prewhitening of random vector x, is exactly by whitening matrix T, has
make the vector after converting
correlation matrix meet
a unit matrix I, second-order statistics between the component after albefaction
independent.
Step 2: initialization population.Gradient formula is utilized to produce population, and the parameter of initialization population, comprise c
1and c
2value, maximum iteration time K
max, number of particles M, blind source separating precision ε, fitness variance threshold values δ, the maximal value of inertia weight and minimum value.
Step 3: the fitness calculating each particle according to formula (3), upgrades individual optimal value and colony's optimal value simultaneously.
Step 4: the average fitness obtaining whole population according to the fitness of each particle.
Step 5: contrasted with average fitness one by one by the fitness of each particle, particle fitness being greater than average fitness is classified as outstanding particle, and being less than is then general particle.Outstanding particle adopts Chaos particle swarm optimization algorithm, obtains optimum solution; General particle adopts the particle cluster algorithm of cloud model to ask optimum solution.
Step 6: the globally optimal solution of the particle of being tried to achieve by two kinds of algorithms compares, and upgrades initial globally optimal solution.
Step 7: judge whether to reach maximum iteration time K
max, if so, then export optimum solution; If not, iterations adds 1, returns step 3.
Step 8: global optimum position to be multiplied with hybrid matrix as separation matrix and to obtain separating mixed matrix, algorithm terminates.
Embodiment 3, Chaos particle swarm optimization algorithm step of the present invention.
Reference
fig. 3, a kind of blind source separation method of the present invention, in the present embodiment, the concrete steps of Chaos particle swarm optimization algorithm of the present invention are:
Step one: read the information being assigned as that a part of population of outstanding particle, these information comprise the initial position of each particle, initial velocity, the initial parameter value of population and the fitness of each particle.
Step 2: by following formula adaptive generation Studying factors c
1, c
2with inertia weight ω,
c
1(t)＝2.5-2*exp(-α|f
avg-f
g|)(6)
c
2(t)＝0.5+2*exp(-α|f
avg-f
g|)(7)
In formula, iter
maxit is maximum evolution number of times.F
goptimal-adaptive degree, f
avgbe the average fitness of population, α is control coefrficient.
Step 3: by position and the speed more new formula of following particle cluster algorithm, upgrade individual optimal value and colony's optimal value.
v
id(t+1)＝ω(t)·v
id(t)+c
1(t)r
1(p
id-x
id(t))+c
2(t)r
2(p
gd-x
id(t))(8)
x
id(t+1)＝x
id(t)+v
id(t+1)(9)
In formula, i=1,2 ..., M, M are the scale of population; Namely total number of particles; D=1,2 ..., N, N are the dimensions of search volume; v
idthe speed of particle i when d dimension; x
idthe position of particle i when d dimension; ω (t) is inertia weight; c
1(t) and c
2t () is Studying factors; p
idand p
gdlocal optimum position and the global optimum position of particle i respectively; r
1and r
2be the random number be distributed on [0,1].
Step 4: (3) calculate the fitness of each particle with the formula, calculates the fitness variance of particle with following formula
In formula
if σ
2< δ, represents and occurs that precocious phenomenon appears in particle, need to carry out chaotic disturbance to particle; Otherwise carry out step 5.
Step 5: adaptive scheduling is carried out to inertia weight ω (t) according to formula (5), and according to formula (8) and formula (9) the more position of new particle and speed.
Step 6: judge whether to reach maximum iteration time K
max, if so, stop iteration, return globally optimal solution W, algorithm terminates; If not, return step 2, continue search.
Embodiment 4, Chaos iteration algorithm step of the present invention.
Reference
fig. 4, a kind of blind source separation method of the present invention, in the present embodiment, the concrete steps of Chaos iteration algorithm of the present invention are:
Step one: the information reading precocious particle, these information comprise the initial position of this particle, the fitness of initial velocity and this particle.
Step 2: for precocious particle, carries out chaos iteration with Logistic equation, produces one group of n-dimensional vector z
1, z
2... z
n, wherein Logistic mapping formula is:
Z
n+1＝μZ
n(1-Z
n)(11)
Z'
k＝(1-β)Z
*+βZ
k(12)
In formula, μ ∈ [3.75,4] is Logistic parameter, Z
nrepresent Chaos Variable, Z
kfor chaos vector when k time, Z'
kfor adding the chaos vector after disturbance, β ∈ [0,1] represents the intensity of disturbance, and n is an integer.Available following formula is to the variable Z not within the scope of this
n∈ (a
i, b
i) carry out coming and going mapping.
cZ
n＝(Z
n-a
i)/(b
i-a
i)(14)
Z
n＝a
i+cZ
n·(b
i-a
i)(15)
Step 3: the fitness value calculating the particle after chaotic disturbance, finds out one group that fitness is maximum, and compares with the fitness without optimal location during chaos optimization, if be better than g
best, just to g
bestupgrade, algorithm terminates; If not, abandon this optimum solution, return step 2.
Embodiment 5, cloud model particle cluster algorithm step of the present invention.
Reference
fig. 5, a kind of blind source separation method of the present invention, in the present embodiment, the concrete steps of cloud model particle cluster algorithm of the present invention are:
Step one: read the information being assigned as that a part of population of general particle, these information comprise the initial position of each particle, initial velocity, the initial parameter value of population and the fitness of each particle.
Step 2: three the mathematical feature { E being set cloud model by following formula
x, E
n, H
e.
E
n＝f
avg'(16)
E
n＝(f
avg'-f
g')÷c
c1(17)
H
e＝E
n÷c
c2(18)
E
n'＝normrnd(E
n,H
e)(19)
Wherein f
avg' be the average fitness of this part population adopting cloud model; f
g' be the optimal-adaptive degree of this part population; c
c1, c
c2for the controling parameters of cloud model.
Step 3: under the restriction of given three mathematical features, generate water dust by X condition cloud generator, concrete generation step is as follows:
Input:{E
x, E
n, H
e, n, x
0// numerical characteristic and water dust number
ouput：{(x
0,μ
c1),…(x
0,μ
cn)}
fori＝1ton
E
n'＝randn(E
n,H
e)
drop(x
0,μ
c)
Step 4: generate adaptable inertia weigh ω according to following formula:
Step 5: according to formula (8) (9) the more speed of new particle and position, more individual extreme value and global extremum, upgrade globally optimal solution, and compared by the fitness with optimal particle when optimizing without cloud model, if be better than g
best, just to g
bestupgrade, algorithm terminates; If not, abandon this optimum solution, return step 4.
The matlab program of the algorithm of embodiments of the invention 2 ~ embodiment 5 is as follows:
clc；clearall；closeall；
function[xm,fv]＝CPSO(M,w,c1,c2,xmax,xmin,Kmax,MaxC,N)
The objective function that/* % is to be optimized: fitness; % number of particles: M; % inertia weight w; % Studying factors: c1, c2; The maximal value of % independent variable search: xmax; The minimum value of % independent variable search: xmin; % maximum iteration time: Kmax; The maximum step number Max:C of % search; The dimension of % particle: N; % objective function obtains argument value during maximal value: xm; % objective function maximal value: fv; */
。
Claims (5)
1. a blind source separation method, has the modeling of blind source separating problem, fitness evaluation and separation matrix to solve three steps:
Described blind source separating problem modeling is under the assumed condition of statistical iteration, is recovered the process of source signal by observation signal; Particular by searching separation matrix, each vector of output matrix obtained after making this matrix and observing the signal matrix obtained be multiplied is separate, then this output matrix is exactly the estimated signal of source signal;
Described fitness evaluation is, after particle initialization, the fitness according to particle is divided into two populations particle, according to formula kurt (the y)=E{y asking fitness
4-3 (E{y
2)
2, obtain the average fitness of all particles, the particle being greater than average fitness adopts Chaos particle swarm optimization algorithm to try to achieve global optimum position; The particle being less than average fitness adopts cloud model particle cluster algorithm to generate the method for inertia weight, tries to achieve global optimum position; Described particle particle cluster algorithm initialization procedure on emulation platform produces at random; Described initialization refers to, first the matrix of stochastic generation n × n dimension is as initially-separate matrix, using the initial global optimum position of each for this matrix column vector as particle, limit in interval in Studying factors, inertia weight and speed, all particles are drawn close to the personal best particle making fitness maximum and are upgraded personal best particle, all personal best particles are arranged as the matrix of a global optimum position, this matrix is the separation matrix of n × n dimension;
Described separation matrix solves and is, fitness is greater than to the particle of average fitness, Chaos particle swarm optimization algorithm is adopted to solve, particularly utilize chaology generate one group with the same number of Chaos Variable of problem control variable to be asked, chaotic disturbance is carried out to control variable, the traversal scope of chaos is transformed into the restriction range of control variable, finally, according to speed and the location updating formula iteration of population, seek the globally optimal solution of problem; Particle fitness being less than to average fitness adopts cloud model particle cluster algorithm to solve, X condition cloud generator is particularly utilized to generate water dust as particle, and adaptive generation inertia weight, the speed of recycling particle cluster algorithm and location updating formula, upgrade the individual extreme value of water dust and global optimum, thus find the position making fitness reach maximum one group of water dust to separate as output; The optimal location that comparison two kinds of optimizing algorithms are obtained obtains global optimum position, and this optimal location is also globally optimal solution, solves the separation matrix of mixed matrix as blind source separating.
2. a kind of blind source separation method according to claim 1, is characterized in that, the mathematical modeling of described blind source separating problem, particularly according to formula X (t)=AS (t)+n (t) and
modeling is carried out, wherein X (t)=[x to the instantaneous mixing of blind source separating and Signal separator
1, x
2..., x
m]
tfor observation signal set, x
i∈ C
mfor each road signal of actual observation, S (t)=[s
1, s
2..., s
n]
t, s
i∈ C
nfor mutual statistical independently source signal, n (t) is m × n dimension additive noise, and A is hybrid matrix, and W is separation matrix.
3. a kind of blind source separation method according to claim 1, is characterized in that, described employing Chaos particle swarm optimization algorithm solves, and is introduce chaos iteration in particle cluster algorithm, particularly according to logical mappings formula Z
n+1=4Z
n(1-Z
n) represent, in formula, Z
nrepresent Chaos Variable, according to chaos principle, chaotic disturbance is added to population, i.e. Z'
k=(1-β) Z
*+ β Z
k, in formula, Z
kfor chaos vector when k time, Z'
kfor adding the chaos vector after disturbance, β ∈ [0,1] represents the intensity of disturbance, adopts self-adaptation value, and at the search initial stage, its value is comparatively large, strengthens the disturbance to solution vector, and along with going deep into of search, β slowly reduces, and is specifically changed to
in formula, n is an integer.
4. a kind of blind source separation method according to claim 1, is characterized in that, described employing cloud model particle cluster algorithm solves, and is the inertia weight being generated population by following formula; E
n=f
avg' E
n=(f
avg'-f
g') ÷ c
c1, H
e=E
n÷ c
c2, E
n'=normrnd (E
n, H
e); Wherein f
avg' be the average fitness of this part population adopting cloud model; f
g' be the optimal-adaptive degree of this part population; c
c1, c
c2for the controling parameters of cloud model.
5., according to the arbitrary described a kind of blind source separation method of Claims 1 to 4, it is characterized in that, described particle cluster algorithm refers to have parameter adaptive particle cluster algorithm, particularly according to formula v
id(t+1)=ω v
id(t)+c
1r
1(p
id-x
id(t))+c
2r
2(p
gd-x
id(t)) and formula x
id(t+1)=x
id(t)+v
id(t+1) the renewal iteration of position and speed is carried out, wherein i=1,2 ..., M, M are the scale of population, namely total number of particles; D=1,2 ..., N, N are the dimensions of search volume; v
idthe speed of particle i when d dimension; x
idthe position of particle i when d dimension; ω is inertia weight; c
1and c
2it is Studying factors; p
idand p
gdlocal optimum position and the global optimum position of particle i respectively; r
1and r
2be the random number be distributed on [0,1], described parameter adaptive, comprise Studying factors c
1, c
2, wherein c
1(t)=2.5-2*exp (-α | f
avg-f
g|), c
2(t)=0.5+2*exp (-α | f
avg-f
g|), f
goptimal-adaptive degree, f
avgbe the average fitness of population, α is control coefrficient; Adaptive generation inertia weight ω,
in formula, iter
maxit is maximum evolution number of times; The fitness variance of each particle is σ
2,
in formula
f
iit is the fitness of i-th particle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510861672.6A CN105404783A (en) | 2015-12-01 | 2015-12-01 | Blind source separation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510861672.6A CN105404783A (en) | 2015-12-01 | 2015-12-01 | Blind source separation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105404783A true CN105404783A (en) | 2016-03-16 |
Family
ID=55470269
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510861672.6A Pending CN105404783A (en) | 2015-12-01 | 2015-12-01 | Blind source separation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105404783A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107356905A (en) * | 2017-08-18 | 2017-11-17 | 华南理工大学 | A kind of visible ray localization method and its system based on chaotic particle swarm optimization |
| CN108535208A (en) * | 2018-03-28 | 2018-09-14 | 广西科技大学 | The control method and control system of gas detecting device |
| CN108615068A (en) * | 2018-03-24 | 2018-10-02 | 西安电子科技大学 | A kind of particle group optimizing method of chaotic disturbance and adaptive inertia weight |
| CN108656117A (en) * | 2018-05-21 | 2018-10-16 | 武汉理工大学 | The mechanical arm spacing track optimizing method of optimal time under a kind of multi-constraint condition |
| CN108899919A (en) * | 2018-08-21 | 2018-11-27 | 江苏方天电力技术有限公司 | A kind of active power distribution network Dynamic reactive power optimization dispatching method based on Haar wavelet transformation |
| CN109059908A (en) * | 2018-06-28 | 2018-12-21 | 中国科学院光电技术研究所 | Disturbance information extracting method in a kind of motion platform photoelectric follow-up inertial sensor signal |
| CN109523488A (en) * | 2018-11-23 | 2019-03-26 | 南京邮电大学 | A kind of radioscopic image noise-reduction method carrying out blind source separating in Anscombe transform domain based on image sequence |
| CN112001256A (en) * | 2020-07-24 | 2020-11-27 | 东风汽车集团有限公司 | Method and system for removing power frequency interference of mixed signal |
| CN113470072A (en) * | 2021-07-06 | 2021-10-01 | 吉林省田车科技有限公司 | Particle swarm target tracking algorithm based on moving particles |
| CN113534790A (en) * | 2021-05-18 | 2021-10-22 | 广西综合交通大数据研究院 | Path planning method and device, electronic equipment and computer readable storage medium |
| CN114172766A (en) * | 2021-12-06 | 2022-03-11 | 东莞理工学院城市学院 | Signal separation method, signal separation device, electronic apparatus, and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102880904A (en) * | 2012-08-24 | 2013-01-16 | 广西南宁推特信息技术有限公司 | Particle swarm optimization algorithm solving method and system for dining recommendation and assignment problems |
| CN103427791A (en) * | 2013-08-08 | 2013-12-04 | 长安大学 | Signal separation method based on particle swarm optimization |
| CN103942612A (en) * | 2014-04-08 | 2014-07-23 | 河海大学 | Cascade reservoir optimal operation method based on adaptive particle swarm optimization algorithm |
| CN104376360A (en) * | 2014-11-11 | 2015-02-25 | 浪潮电子信息产业股份有限公司 | Novel chaotic particle swarm optimization algorithm |
-
2015
- 2015-12-01 CN CN201510861672.6A patent/CN105404783A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102880904A (en) * | 2012-08-24 | 2013-01-16 | 广西南宁推特信息技术有限公司 | Particle swarm optimization algorithm solving method and system for dining recommendation and assignment problems |
| CN103427791A (en) * | 2013-08-08 | 2013-12-04 | 长安大学 | Signal separation method based on particle swarm optimization |
| CN103942612A (en) * | 2014-04-08 | 2014-07-23 | 河海大学 | Cascade reservoir optimal operation method based on adaptive particle swarm optimization algorithm |
| CN104376360A (en) * | 2014-11-11 | 2015-02-25 | 浪潮电子信息产业股份有限公司 | Novel chaotic particle swarm optimization algorithm |
Non-Patent Citations (3)
| Title |
|---|
| 彭安洪 等: "一种采用云自适应粒子群算法的盲源分离", 《计算机仿真》 * |
| 杨汉华: "参数自适应混沌粒子群算法在盲源分离中的应用", 《微电子学与计算机》 * |
| 谢春明 等: "混沌粒子群算法的盲源信号分离仿真研究", 《计算机应用与软件》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107356905B (en) * | 2017-08-18 | 2023-04-28 | 华南理工大学 | Visible light positioning method and system based on chaotic particle swarm optimization |
| CN107356905A (en) * | 2017-08-18 | 2017-11-17 | 华南理工大学 | A kind of visible ray localization method and its system based on chaotic particle swarm optimization |
| CN108615068A (en) * | 2018-03-24 | 2018-10-02 | 西安电子科技大学 | A kind of particle group optimizing method of chaotic disturbance and adaptive inertia weight |
| CN108615068B (en) * | 2018-03-24 | 2020-07-14 | 西安电子科技大学 | Particle swarm optimization method based on chaotic disturbance and adaptive inertial weight |
| CN108535208A (en) * | 2018-03-28 | 2018-09-14 | 广西科技大学 | The control method and control system of gas detecting device |
| CN108656117A (en) * | 2018-05-21 | 2018-10-16 | 武汉理工大学 | The mechanical arm spacing track optimizing method of optimal time under a kind of multi-constraint condition |
| CN109059908A (en) * | 2018-06-28 | 2018-12-21 | 中国科学院光电技术研究所 | Disturbance information extracting method in a kind of motion platform photoelectric follow-up inertial sensor signal |
| CN108899919A (en) * | 2018-08-21 | 2018-11-27 | 江苏方天电力技术有限公司 | A kind of active power distribution network Dynamic reactive power optimization dispatching method based on Haar wavelet transformation |
| CN109523488A (en) * | 2018-11-23 | 2019-03-26 | 南京邮电大学 | A kind of radioscopic image noise-reduction method carrying out blind source separating in Anscombe transform domain based on image sequence |
| CN112001256A (en) * | 2020-07-24 | 2020-11-27 | 东风汽车集团有限公司 | Method and system for removing power frequency interference of mixed signal |
| CN112001256B (en) * | 2020-07-24 | 2023-05-26 | 东风汽车集团有限公司 | Mixed signal power frequency interference removing method and system |
| CN113534790A (en) * | 2021-05-18 | 2021-10-22 | 广西综合交通大数据研究院 | Path planning method and device, electronic equipment and computer readable storage medium |
| CN113470072A (en) * | 2021-07-06 | 2021-10-01 | 吉林省田车科技有限公司 | Particle swarm target tracking algorithm based on moving particles |
| CN113470072B (en) * | 2021-07-06 | 2022-09-13 | 吉林省田车科技有限公司 | Particle swarm target tracking algorithm based on moving particles |
| CN114172766A (en) * | 2021-12-06 | 2022-03-11 | 东莞理工学院城市学院 | Signal separation method, signal separation device, electronic apparatus, and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105404783A (en) | Blind source separation method | |
| Jiang et al. | Deep-learning-based joint resource scheduling algorithms for hybrid MEC networks | |
| Dogo et al. | A comparative analysis of gradient descent-based optimization algorithms on convolutional neural networks | |
| Yaghini et al. | A hybrid algorithm for artificial neural network training | |
| Sun et al. | A hybrid co-evolutionary cultural algorithm based on particle swarm optimization for solving global optimization problems | |
| Pehlivanoglu | A new particle swarm optimization method enhanced with a periodic mutation strategy and neural networks | |
| Li et al. | Hybrid optimization algorithm based on chaos, cloud and particle swarm optimization algorithm | |
| Wu et al. | A novel sparrow search algorithm for the traveling salesman problem | |
| Lin et al. | Machine learning templates for QCD factorization in the search for physics beyond the standard model | |
| CN108776820A (en) | It is a kind of to utilize the improved random forest integrated approach of width neural network | |
| CN106203507A (en) | A kind of k means clustering method improved based on Distributed Computing Platform | |
| Aladag et al. | A new seasonal fuzzy time series method based on the multiplicative neuron model and SARIMA | |
| Zhang et al. | A novel fuzzy hybrid quantum artificial immune clustering algorithm based on cloud model | |
| CN116402002B (en) | Multi-target layered reinforcement learning method for chip layout problem | |
| Lin et al. | K-means optimization clustering algorithm based on particle swarm optimization and multiclass merging | |
| Fei et al. | Research on data mining algorithm based on neural network and particle swarm optimization | |
| Banerjee et al. | Boosting exploration in actor-critic algorithms by incentivizing plausible novel states | |
| Wang et al. | The study of K-means based on hybrid SA-PSO algorithm | |
| Esmin et al. | Data clustering using hybrid particle swarm optimization | |
| Wei et al. | An energy-efficient partition-based framework with continuous ant colony optimization for target tracking in mobile sensor networks | |
| Song et al. | Prediction for chaotic time series of optimized BP neural network based on modified PSO | |
| Wang et al. | A spark-based artificial bee colony algorithm for large-scale data clustering | |
| Yaghini et al. | HIOPGA: a new hybrid metaheuristic algorithm to train feedforward neural networks for prediction | |
| Huo et al. | An improved artificial bee colony algorithm for numerical functions | |
| Zhang et al. | Improved salp swarm algorithm based on gravitational search and multi-leader search strategies |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20160316 |