CN107346423A - The face identification method of autoassociative memories based on cell neural network - Google Patents

The face identification method of autoassociative memories based on cell neural network Download PDF

Info

Publication number
CN107346423A
CN107346423A CN201710526609.6A CN201710526609A CN107346423A CN 107346423 A CN107346423 A CN 107346423A CN 201710526609 A CN201710526609 A CN 201710526609A CN 107346423 A CN107346423 A CN 107346423A
Authority
CN
China
Prior art keywords
face picture
mrow
value
face
formula
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
CN201710526609.6A
Other languages
Chinese (zh)
Inventor
韩琦
邓世琴
刘晋
翁腾飞
熊思斯
吴政阳
谯自强
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.)
Chongqing University of Science and Technology
Original Assignee
Chongqing 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 Chongqing University of Science and Technology filed Critical Chongqing University of Science and Technology
Priority to CN201710526609.6A priority Critical patent/CN107346423A/en
Publication of CN107346423A publication Critical patent/CN107346423A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/002—Biomolecular computers, i.e. using biomolecules, proteins, cells
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Organic Chemistry (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Chemical & Material Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a kind of face identification method of the autoassociative memories based on cell neural network, including S1：Random acquisition m width face pictures, and be numbered；S2：By setting binary map luminance threshold, the m width face picture that step S1 is obtained is handled as two-value face picture respectively, obtains the autoassociative memories input matrix and output matrix of two-value face picture；S3：Build the cell neural network face picture identification model framework containing unknown model parameters；S4：Unknown model parameters, determine final cell neutral net face picture identification model；S5：Based on autoassociative memories criterion, matching is identified to any face picture.Beneficial effect：Digitization preserves, and safety coefficient is high；Recognition effect is good, and recognition efficiency is high；Matching result is with a high credibility.

Description

The face identification method of autoassociative memories based on cell neural network
Technical field
The present invention relates to image identification technical field, specifically a kind of autoassociative memories based on cell neural network Face identification method.
Background technology
With the development in big data epoch, there is identity information verification, such as recognition of face in people during trip.It is logical Identification is crossed, realizes authentication, improves the security performance of system, confirms different subscriber identity informations.
When carrying out face information verification, the face information that is necessarily preserved including database and etc. face letter to be verified , in the prior art, following defect be present in breath, the face information preserved for database：
First：Face information often by the way of directly storing, so causes identity information easily to reveal, safety system Number is low, once leakage, is easy for being replicated, poor reliability.
Second：, it is necessary to which substantial amounts of image data in database is transferred and compared in face picture identification process, adjust Take picture time to grow, cause recognition efficiency low, mistake easily occur.
3rd：During picture recognition, when finding similar picture, in the absence of picture match, poor reliability, it is easier to occur Check mistake.
For drawbacks described above, it is necessary to a kind of face picture recognition methods is proposed, to meet the needs of people.
The content of the invention
In view of the above-mentioned problems, the invention provides a kind of recognition of face side of the autoassociative memories based on cell neural network Method, face picture is subjected to digitization preservation, combining association memory and Cellular Neural Networks, realizes that face picture identifies, Safety coefficient is high, good reliability, and recognition efficiency is high.
To reach above-mentioned purpose, the concrete technical scheme that the present invention uses is as follows：
A kind of face identification method of the autoassociative memories based on cell neural network, its key are to include following step Suddenly：
S1：The face picture of z people in crowd is acquired at random, m=w*z width face picture is obtained wherein, w is Positive integer, and be numbered to collecting m width face pictures；
S2：By setting binary map luminance threshold, the m width face picture that step S1 is obtained is handled as two-value face respectively Picture, obtain the autoassociative memories input matrix and output matrix being made up of m two-value face pictures；
S3：Build the cell neural network face picture identification model framework containing unknown model parameters；
S4：What face picture autoassociative memories input matrix, output matrix and the step S3 obtained according to step S2 was obtained Cell neural network face picture identification model framework, calculate the Unknown Model ginseng of cell neural network face picture identification model Number, it is final to determine cell neural network face picture identification model；
S5：Based on autoassociative memories criterion, matching is identified to any face picture.
By above-mentioned design, face picture is realized that digitization preserves, passes through cell neural network and associative memory method Cell neural network face picture identification model is established, image data identification process is changed into the identification process of numerical data, Recognition time is shortened, avoid picture transfers process and circulation comparison procedure, and this method also carries out final to recognition result Matching treatment, improve identification certainty.
Further describe, the binary map luminance threshold K ∈ { 0,1,2,3 ..., 255 }；For different face pictures Database, different binary map luminance thresholds can be set, improve the reliability and safety coefficient of face picture database.
Each width two-value face picture is arranged to the picture for including N row M row pixels, pixel total number be n=N × M；
When being preserved to face picture, if the picture got is irregular picture, by the irregular picture Minimum rectangle is carried out, i.e., irregular picture need to be completely included using rectangle frame, the rectangle frame, the blank being had more in rectangle frame Part people can be filled using customized pattern.
If the output matrix of two-value face picture associative memory is：O=(α1,α2,…,αi,…,αm),I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, αiRepresent the i-th width two-value face figure The output vector of all pixels composition in piece,Represent the output of j-th of pixel in the i-th width two-value face picture Value；
If the input matrix of two-value face picture associative memory is：I=(U1,U2,...,Ui,…,Um),I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, UiRepresent in the i-th width two-value face picture The input vector of all pixel compositions,Represent the input value of j-th of pixel in the i-th width two-value face picture.
Further describe, step S3 builds the cell neural network face picture identification model containing unknown model parameters Framework concretely comprises the following steps：
Build cell neural network face picture identification model basic framework：
Wherein, x=(x1,x2,…,xi,…,xn)T, i ∈ { 1,2 ..., n }；
Input vector U=(u1,u2..., ui,…,un)T, i ∈ { 1,2 ..., n }；
Given coefficient C=diag (c1,c2,…,ci,…,cn), i ∈ { 1,2 ..., n }；
Activation primitive f (x)=(f (x1),…,f(xi),…,f(xn))T；
Offset vector V=(v1,v2,…,vi,…,vn)T, i ∈ { 1,2 ..., n }；
Wherein given coefficient C is artificial given；Coefficient matrices A, coefficient matrix D and offset V are unknown model parameters；
In formula (1), coefficient matrices A=(aij)n×nIt is made up of following square formation：
Wherein,
In formula (1), coefficient matrix D=(dij)n×nIt is made up of following square formation：
Wherein,
Make α=(α1,α2,…,αi,…,αn)T∈Υn={ x=(x1,x2,…,xi,…,xn)T∈Rn|xi=1 or xi =-1 }；
Make C (α)={ y=(y1,y2,…,yi,…,yn)T∈Rn|yiαi＞ 1 }；
Therefore, the cell neural network face picture identification model framework containing unknown model parameters is obtained by formula (1)：
Further describe, step S4's concretely comprises the following steps：
S41：The formula (2) that step S3 is obtained can be write as following form：
Make xi(0)=0,
If (i)Then formula (3) converges to a positive stabilization equalization point, and this The value of equalization point is more than 1；
(ii) ifThen formula (3) converges to a negative stable equilibrium point, and this The value of individual equalization point is less than -1；
Then obtain the first inference：
Orderci=constant, i ∈ 1,2 ..., n }；
Work as αiWhen=1, formula (3) converges to a positive stabilization equalization point, and the value of this equalization point is more than 1；
Work as αiWhen=- 1, formula (3) converges to a negative stable equilibrium point, and the value of this equalization point is less than -1；
Created symbol Λ, makes
Wherein, λi＞ 0；Then
If l ∈ { 1,2 ..., m }, q ∈ { 1,2 ..., N }；
Input parameter：
Output parameter：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：Δ=Λ ' Γ '；
According to first inference, obtain：
The Δ (5) of Φ '+V '=0.55
AO+DU+V "=Λ O (6)
Formula (5) is converted to：
The Δ (7) of Ω LA+V '=0.55
It can be obtained according to formula (7)：
LA=pinv (Ω) (0.55 Δ-V ') (8)
Wherein, the pseudoinverse of pinv () representing matrix；
Formula (6) is converted to
Ξ LD+ Φ '+V '=Δ (9)
Therefore, can be obtained by formula (9)
LD=pinv (Ξ) (Δ-Ω LA-V ') (10)
S42：In autoassociative memories criterion, input matrix be present and be equal to output matrix, i.e.,
I=O (11)
Step S2 is obtained into the autoassociative memories output matrix O=(α of two-value face picture1,α2,…,αm) and input matrix I=(U1,U2,…,Um) matrix Ω and Ξ are converted into, and bring formula (8) into and formula (10) obtains output parameter：
LA=pinv (Ω) (0.55 Δ-V ')；
Input parameter：
LD=pinv (Ξ) (Δ-Ω LA-V ')；
S43：By the output parameter LA that step S42 is obtained and the coefficient matrices A that input parameter LD is changed into step S3 and Coefficient matrix D；
Offset vector V is obtained according to formula (4)；
S44：The given coefficient C of setting, and the Unknown Model coefficient that step S43 is obtained is brought into formula (2) and obtains cell Neutral net face picture identification model.
Further describe, step S5 particular content is：
S51：Any face picture is obtained, carrying out binaryzation to face picture obtains the autoassociative memories of two-value face picture Input matrix；
S52：The obtained input matrixes of step S51 are updated into the cell neural network face picture that step S4 obtains to identify In model, model output matrix is obtained；
S53：The model output matrix that the input matrix that step S51 is obtained obtains with step S52 is respectively to face picture Matched, obtain face picture the match is successful that rate is H；
S54：Judge face picture the match is successful whether rate H is more than match settings value h, h=0~1；If so, for matching into Work(, otherwise it fails to match.
Beneficial effects of the present invention：Autoassociative memories and Cellular Neural Networks are combined, face picture is converted Preserved into series of parameters, identity information is face picture, and checking reliability is high, and concealed to the preserving type of picture Property it is strong, safety coefficient is high, effectively prevents that people's identity information is compromised；Form by picture through model conversation into parameter, simply Convenient, practicality is good, and picture recognition effect is good, good to face picture protecting effect, for irregular picture, carries out minimum square Shapeization is filled, and improves the feasibility of this method.
Brief description of the drawings
Fig. 1 is the present inventor's face recognition method flow chart；
Fig. 2 is cell neural network face picture identification model unknown model parameters resolution principle figure of the present invention.
Embodiment
The embodiment and operation principle of the present invention are described in further detail below in conjunction with the accompanying drawings.
It will be seen from figure 1 that a kind of face identification method of the autoassociative memories based on cell neural network, including it is following Step：
S1：The face picture of z people in crowd is acquired at random, m=w*z width face picture is obtained wherein, w is Positive integer, and be numbered to collecting m width face pictures；
S2：By setting binary map luminance threshold, the m width face picture that step S1 is obtained is handled as two-value face respectively Picture, obtain the autoassociative memories input matrix and output matrix of two-value face picture；
The binary map luminance threshold K ∈ 0,1,2,3 ..., 255 }；
For different face picture databases, different binary map luminance thresholds is set, can improve face picture can By property.
For irregular picture, using minimum rectangle frame, irregular picture is filled to the minimum rectangle inframe, it is right In the uncovered part of minimum rectangle frame, covered using white.Make the picture that finally preserves rectangular.
Each width two-value face picture is arranged to the picture for including N row M row pixels, pixel total number be n=N × M；N × M of formation picture is regarded as a matrix, then can obtain：
The output matrix of two-value face picture associative memory is：O=(α1,α2,…,αi,…,αm),I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, αiRepresent the i-th width two-value face figure The output vector of all pixels composition in piece,Represent the output of j-th of pixel in the i-th width two-value face picture Value；
The input matrix of two-value face picture associative memory is：I=(U1,U2,…,Ui,…,Um),I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, UiRepresent in the i-th width two-value face picture The input vector of all pixel compositions,Represent the input value of j-th of pixel in the i-th width two-value face picture.
S3：Build the cell neural network face picture identification model framework containing unknown model parameters；
Step S3 builds the specific step of the cell neural network face picture identification model framework containing unknown model parameters Suddenly it is：
Build cell neural network face picture identification model basic framework：
Wherein, x=(x1,x2,…,xi,…,xn)T, i ∈ { 1,2 ..., n }；
Input vector U=(u1,u2…,ui,…,un)T, i ∈ { 1,2 ..., n }；
Given coefficient C=diag (c1,c2,…,ci,…,cn), i ∈ { 1,2 ..., n }；
Activation primitive f (x)=(f (x1),…,f(xi),…,f(xn))T；
Offset vector V=(v1,v2,…,vi,…,vn)T, i ∈ { 1,2 ..., n }；
Wherein given coefficient C is artificial given；Coefficient matrices A, coefficient matrix D and offset V are unknown model parameters；
In formulaIn, coefficient matrices A=(aij)n×nIt is made up of following square formation：
Wherein,
In formulaIn, coefficient matrix D=(dij)n×nIt is made up of following square formation：
Wherein,
Make α=(α 1, α 2 ..., αi,…,αn)T∈Υn={ x=(x1,x2,…,xi,…,xn)T∈Rn|xi=1 or xi =-1 }；
Make C (α)={ y=(y1,y2,…,yi,…,yn)T∈Rn|yiαi＞ 1 }；
Therefore, by formulaObtain the cell neural network people containing unknown model parameters Face picture recognition model framework：
S4：What face picture autoassociative memories input matrix, output matrix and the step S3 obtained according to step S2 was obtained Cell neural network face picture identification model framework, calculate the Unknown Model ginseng of cell neural network face picture identification model Number, it is final to determine cell neural network face picture identification model；
Step S4's concretely comprises the following steps：
S41：The formula that step S3 is obtainedIt can be write as following form：
Make xi(0)=0,
If (i)Then formulaConvergence To a positive stabilization equalization point, and the value of this equalization point is more than 1；
(ii) ifThen formulaReceive Hold back to a negative stable equilibrium point, and the value of this equalization point is less than -1；
Then obtain the first inference：
Orderci=constant, i ∈ 1,2 ..., n }；
Work as αiWhen=1, formulaConverge to a positive stabilization equalization point, and this The value of individual equalization point is more than 1；
Work as αiWhen=- 1, formulaA negative stable equilibrium point is converged to, and The value of this equalization point is less than -1；
Created symbol Λ, makes
Wherein, λi＞ 0；Then
If l ∈ { 1,2 ..., m }, q ∈ { 1,2 ..., N }；
Input parameter：
Output parameter：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：Δ=Λ ' Γ '；
According to first inference, obtain：
The Δ AO+DU+V " of Φ '+V '=0.55=Λ O
The Δ of formula Φ '+V '=0.55 is converted to：
The Δ of Ω LA+V '=0.55
It can be obtained according to the Δ of formula Ω LA+V '=0.55：
LA=pinv (Ω) (0.55 Δ-V ')
Wherein, the pseudoinverse of pinv () representing matrix；
Formula AO+DU+V "=Λ O are converted to
Ξ LD+ Φ '+V '=Δ
Therefore, can be obtained by formula Ξ LD+ Φ '+V '=Δ
LD=pinv (Ξ) (Δ-Ω LA-V ')
S42：In autoassociative memories criterion, input matrix be present and be equal to output matrix, i.e.,
I=O
Step S2 is obtained into the autoassociative memories output matrix O=(α of two-value face picture1,α2,…,αm) and input matrix I=(U1,U2,…,Um) matrix Ω and Ξ are converted into, and bring formula LA=pinv (Ω) (0.55 Δ-V ') and formula LD=into Pinv (Ξ) (Δ-Ω LA-V ') obtains output parameter：
LA=pinv (Ω) (0.55 Δ-V ')；
Input parameter：
LD=pinv (Ξ) (Δ-Ω LA-V ')；
S43：By the output parameter LA that step S42 is obtained and the coefficient matrices A that input parameter LD is changed into step S3 and Coefficient matrix D；
According to formulaObtain offset vector V；
S44：The given coefficient C of setting, and bring the Unknown Model coefficient that step S43 is obtained into formulaIn obtain cell neural network face picture identification model.
S5：Based on autoassociative memories criterion, matching is identified to any face picture.
Step S5 particular content is：
S51：Any face picture is obtained, carrying out binaryzation to face picture obtains the autoassociative memories of two-value face picture Input matrix；
S52：The obtained input matrixes of step S51 are updated into the cell neural network face picture that step S4 obtains to identify In model, model output matrix is obtained；
S53：The model output matrix that the input matrix that step S51 is obtained obtains with step S52 is respectively to face picture Matched, obtain face picture the match is successful that rate is H；
S54：Judge face picture the match is successful whether rate H is more than match settings value h, h=0~1；If so, for matching into Work(, otherwise it fails to match.
It should be pointed out that it is limitation of the present invention that described above, which is not, the present invention is also not limited to the example above, What those skilled in the art were made in the essential scope of the present invention changes, is modified, adds or replaces, and also should Belong to protection scope of the present invention.

Claims (5)

1. a kind of face identification method of the autoassociative memories based on cell neural network, it is characterised in that comprise the following steps：
S1：The face picture of z people in crowd is acquired at random, m=w*z width face pictures are obtained, wherein, w is just Integer, and be numbered to collecting m width face pictures；
S2：By setting binary map luminance threshold, the m width face picture that step S1 is obtained is handled as two-value face figure respectively Piece, obtain the autoassociative memories input matrix and output matrix of two-value face picture；
S3：Build the cell neural network face picture identification model framework containing unknown model parameters；
S4：The cell that face picture autoassociative memories input matrix, output matrix and the step S3 obtained according to step S2 is obtained Neutral net face picture identification model framework, the unknown model parameters of cell neural network face picture identification model are calculated, It is final to determine cell neural network face picture identification model；
S5：Based on autoassociative memories criterion, matching is identified to any face picture.
2. the face identification method of the autoassociative memories according to claim 1 based on cell neural network, its feature exist In the binary map luminance threshold K ∈ { 0,1,2,3 ..., 255 }；
Each width two-value face picture is arranged to the picture for including N row M row pixels, pixel total number is n=N × M；
If the output matrix of two-value face picture associative memory is：O=(α1,α2,…,αi,…,αm),I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, αiRepresent the i-th width two-value face figure The output vector of all pixels composition in piece,Represent the output of j-th of pixel in the i-th width two-value face picture Value；
If the input matrix of two-value face picture associative memory is：I=(U1,U2,…,Ui,…,Um), I ∈ { 1,2 ..., m }, j ∈ { 1,2 ..., n }, UiRepresent the inputs of all pixel compositions in the i-th width two-value face picture to Amount,Represent the input value of j-th of pixel in the i-th width two-value face picture.
3. the face identification method of the autoassociative memories according to claim 2 based on cell neural network, its feature exist Concretely comprising the following steps for the cell neural network face picture identification model framework containing unknown model parameters is built in step S3：
Build cell neural network face picture identification model basic framework：
<mrow> <mover> <mi>x</mi> <mo>&amp;CenterDot;</mo> </mover> <mo>=</mo> <mo>-</mo> <mi>Cx</mi> <mo>+</mo> <mi>Af</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>DU</mi> <mo>+</mo> <mi>V</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>
Wherein, x=(x1,x2,…,xi,…,xn)T, i ∈ { 1,2 ..., n }；
Input vector U=(u1,u2,…ui,…,un)T, i ∈ { 1,2 ..., n }；
Given coefficient C=diag (c1,c2,…,ci,…,cn), i ∈ { 1,2 ..., n }；
Activation primitive f (x)=(f (x1),…,f(xi),…,f(xn))T；
Offset vector V=(v1,v2,…,vi,…,vn)T, i ∈ { 1,2 ..., n }；
Wherein given coefficient C is artificial given；Coefficient matrices A, coefficient matrix D and offset V are unknown model parameters；
In formula (1), coefficient matrices A=(aij)n×nIt is made up of following square formation：
Wherein,
In formula (1), coefficient matrix D=(dij)n×nIt is made up of following square formation：
Wherein,
Make α=(α1,α2,…,αi,…,αn)T∈Υn={ x=(x1,x2,…,xi,…,xn)T∈Rn|xi=1orxi=-1 }；
Make C (α)={ y=(y1,y2,…,yi,…,yn)T∈Rn|yiαi＞ 1 }；
Therefore, the cell neural network face picture identification model framework containing unknown model parameters is obtained by formula (1)：
<mrow> <mover> <mi>x</mi> <mo>&amp;CenterDot;</mo> </mover> <mo>=</mo> <mo>-</mo> <mi>Cx</mi> <mo>+</mo> <mi>A&amp;alpha;</mi> <mo>+</mo> <mi>DU</mi> <mo>+</mo> <mi>V</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>
4. the face identification method of the autoassociative memories according to claim 3 based on cell neural network, its feature exist In concretely comprising the following steps for step S4：
S41：The formula (2) that step S3 is obtained can be write as following form：
<mrow> <msub> <mover> <mi>x</mi> <mo>&amp;CenterDot;</mo> </mover> <mi>i</mi> </msub> <mo>=</mo> <mo>-</mo> <msub> <mi>c</mi> <mi>i</mi> </msub> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>+</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>a</mi> <mi>ij</mi> </msub> <msub> <mi>&amp;alpha;</mi> <mi>j</mi> </msub> <mo>+</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>d</mi> <mi>ij</mi> </msub> <msub> <mi>u</mi> <mi>j</mi> </msub> <mo>+</mo> <msub> <mi>v</mi> <mi>i</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow>
Make xi(0)=0,
If (i)Then formula (3) converges to a positive stabilization equalization point, and this is balanced The value of point is more than 1；
(ii) ifThen formula (3) converges to a negative stable equilibrium point, and this is flat The value of weighing apparatus point is less than -1；
Then obtain the first inference：
Orderci=constant, i ∈ 1,2 ..., n }；
Work as αiWhen=1, formula (3) converges to a positive stabilization equalization point, and the value of this equalization point is more than 1；
Work as αiWhen=- 1, formula (3) converges to a negative stable equilibrium point, and the value of this equalization point is less than -1；
Created symbol Λ, makes
Wherein, λi＞ 0；Then
If l ∈ 1,2 ..., m }, q ∈ 1,2 ..., N }；
Input parameter：
Output parameter：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：
Order：Δ=Λ ' Γ '；
According to first inference, obtain：
<mrow> <mi>V</mi> <mo>=</mo> <mn>0.1</mn> <mo>&amp;times;</mo> <mi>&amp;Lambda;</mi> <mo>&amp;times;</mo> <mfrac> <mrow> <msup> <mi>&amp;alpha;</mi> <mn>1</mn> </msup> <mo>+</mo> <msup> <mi>&amp;alpha;</mi> <mn>2</mn> </msup> <mo>+</mo> <mo>...</mo> <mo>+</mo> <msup> <mi>&amp;alpha;</mi> <mi>i</mi> </msup> <mo>+</mo> <mo>...</mo> <mo>+</mo> <msup> <mi>&amp;alpha;</mi> <mi>m</mi> </msup> </mrow> <mi>m</mi> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> </mrow>
The Δ (5) of Φ '+V '=0.55
AO+DU+V "=Λ O (6)
Formula (5) is converted to：
The Δ (7) of Ω LA+V '=0.55
It can be obtained according to formula (7)：
LA=pinv (Ω) (0.55 Δ-V ') (8)
Wherein, the pseudoinverse of pinv () representing matrix；
Formula (6) is converted to
Ξ LD+ Φ '+V '=Δ (9)
Therefore, can be obtained by formula (9)
LD=pinv (Ξ) (Δ-Ω LA-V ') (10)
S42：In autoassociative memories criterion, input matrix be present and be equal to output matrix, i.e.,
I=O (11)
Step S2 is obtained into the autoassociative memories output matrix O=(α of two-value face picture1,α2,…,αm) and input matrix I= (U1,U2,…,Um) matrix Ω and Ξ are converted into, and bring formula (8) into and formula (10) obtains output parameter：
LA=pinv (Ω) (0.55 Δ-V ')；
Input parameter：
LD=pinv (Ξ) (Δ-Ω LA-V ')；
S43：By the output parameter LA that step S42 is obtained and coefficient matrices A and coefficient that input parameter LD is changed into step S3 Matrix D；
Offset vector V is obtained according to formula (4)；
S44：The given coefficient C of setting, and the Unknown Model coefficient that step S43 is obtained is brought into formula (2) and obtains cellular neural Network face picture recognition model.
5. the face identification method of the autoassociative memories based on cell neural network according to claim 1 or 2 or 3 or 4, It is characterized in that step S5 particular content is：
S51：Any face picture is obtained, binaryzation is carried out to face picture and obtains the autoassociative memories input of two-value face picture Matrix；
S52：The obtained input matrixes of step S51 are updated to the cell neural network face picture identification model that step S4 obtains In, obtain model output matrix；
S53：The model output matrix that the input matrix that step S51 is obtained obtains with step S52 is carried out to face picture respectively Matching, obtains face picture the match is successful that rate is H；
S54：Judge face picture the match is successful whether rate H is more than match settings value h, h=0~1；If so, for the match is successful, it is no Then it fails to match.
CN201710526609.6A 2017-06-30 2017-06-30 The face identification method of autoassociative memories based on cell neural network Pending CN107346423A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710526609.6A CN107346423A (en) 2017-06-30 2017-06-30 The face identification method of autoassociative memories based on cell neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710526609.6A CN107346423A (en) 2017-06-30 2017-06-30 The face identification method of autoassociative memories based on cell neural network

Publications (1)

Publication Number Publication Date
CN107346423A true CN107346423A (en) 2017-11-14

Family

ID=60257927

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710526609.6A Pending CN107346423A (en) 2017-06-30 2017-06-30 The face identification method of autoassociative memories based on cell neural network

Country Status (1)

Country Link
CN (1) CN107346423A (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7996156B2 (en) * 2002-03-07 2011-08-09 The United States Of America As Represented By The Secretary, Department Of Health And Human Services Methods for predicting properties of molecules
CN102194106A (en) * 2011-05-11 2011-09-21 西安理工大学 Human face recognition method used in door access system
CN103310200A (en) * 2013-06-25 2013-09-18 郑州吉瑞特电子科技有限公司 Face recognition method
CN103714351A (en) * 2013-12-18 2014-04-09 五邑大学 Depth self learning-based facial beauty predicting method
CN104484803A (en) * 2014-11-24 2015-04-01 苏州福丰科技有限公司 Mobile phone payment method employing three-dimensional human face recognition based on neural network
CN106295530A (en) * 2016-07-29 2017-01-04 北京小米移动软件有限公司 Face identification method and device
WO2017070858A1 (en) * 2015-10-28 2017-05-04 Beijing Sensetime Technology Development Co., Ltd A method and a system for face recognition

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7996156B2 (en) * 2002-03-07 2011-08-09 The United States Of America As Represented By The Secretary, Department Of Health And Human Services Methods for predicting properties of molecules
CN102194106A (en) * 2011-05-11 2011-09-21 西安理工大学 Human face recognition method used in door access system
CN103310200A (en) * 2013-06-25 2013-09-18 郑州吉瑞特电子科技有限公司 Face recognition method
CN103714351A (en) * 2013-12-18 2014-04-09 五邑大学 Depth self learning-based facial beauty predicting method
CN104484803A (en) * 2014-11-24 2015-04-01 苏州福丰科技有限公司 Mobile phone payment method employing three-dimensional human face recognition based on neural network
WO2017070858A1 (en) * 2015-10-28 2017-05-04 Beijing Sensetime Technology Development Co., Ltd A method and a system for face recognition
CN106295530A (en) * 2016-07-29 2017-01-04 北京小米移动软件有限公司 Face identification method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
韩琦: "神经网络的稳定性及其在联想记忆中的应用研究", 《中国博士学位论文全文数据库信息科技辑》 *

Similar Documents

Publication Publication Date Title
CN109377452B (en) Face image restoration method based on VAE and generation type countermeasure network
CN103745104B (en) A kind of method of marking examination papers based on augmented reality
CN107767413A (en) A kind of image depth estimation method based on convolutional neural networks
CN107944282B (en) A kind of reversible information hidden method based on asymmetric prediction histogram of error modification
CN110211035A (en) Merge the image super-resolution method of the deep neural network of mutual information
WO2020143316A1 (en) Certificate image extraction method and terminal device
CN105404817B (en) It is based on（k, n）The user-friendly visual secret sharing scheme method of thresholding
CN103886235B (en) Face image biological key generating method
CN108510451A (en) A method of the reconstruction car plate based on the double-deck convolutional neural networks
CN107392865A (en) A kind of restored method of facial image
CN108537747A (en) A kind of image repair method based on the convolutional neural networks with symmetrical parallel link
CN109902613A (en) A kind of human body feature extraction method based on transfer learning and image enhancement
CN104794726B (en) A kind of underwater picture Parallel segmentation method and device
CN109426859A (en) Neural metwork training system, method and computer readable storage medium
CN108230308A (en) A kind of handbill proofreading method, device and equipment
CN111191704A (en) Foundation cloud classification method based on task graph convolutional network
CN111144407A (en) Target detection method, system, device and readable storage medium
CN107346423A (en) The face identification method of autoassociative memories based on cell neural network
CN107016684A (en) A kind of super-pixel rapid generation of Polarimetric SAR Image
CN117395067A (en) User data privacy protection system and method for Bayesian robust federal learning
CN107330404A (en) Personal identification method based on cell neural network autoassociative memories model
CN115115552B (en) Image correction model training method, image correction device and computer equipment
CN107154027A (en) Compensation method and device that a kind of fault image restores
CN110175509A (en) A kind of round-the-clock eye circumference recognition methods based on cascade super-resolution
US20220207661A1 (en) Image Descattering Method Based on Iterative Optimization of Atmospheric Transmission Matrix

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20171114

RJ01 Rejection of invention patent application after publication