CN102298704A - Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF) - Google Patents

Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF) Download PDF

Info

Publication number
CN102298704A
CN102298704A CN2011101087757A CN201110108775A CN102298704A CN 102298704 A CN102298704 A CN 102298704A CN 2011101087757 A CN2011101087757 A CN 2011101087757A CN 201110108775 A CN201110108775 A CN 201110108775A CN 102298704 A CN102298704 A CN 102298704A
Authority
CN
China
Prior art keywords
sample
increment
classification
weighting
establish
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
CN2011101087757A
Other languages
Chinese (zh)
Inventor
金连文
刘岗
丁凯
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN2011101087757A priority Critical patent/CN102298704A/en
Publication of CN102298704A publication Critical patent/CN102298704A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Character Discrimination (AREA)

Abstract

The invention provides a writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF). The method is characterized by utilizing an increment sample with the writing style of the specific user to dynamically update an MQDF recognition model to ensure the updated MQDF recognition model to adapt to the writing style of the specific user, thus achieving the effect of improving the recognition rate of the specific user. The method has the following beneficial effects: the WIMQDF algorithm is put forward by creatively combining the weighing incremental learning mechanism with the MQDF classification algorithm used in Chinese character recognition and the WIMQDF is applied to the writer self-adaptation field based on handwritten Chinese character recognition, thus solving the problem that the handwritten Chinese character recognition engine has low recognition rate for the specific user; and by utilizing the method provided by the invention, the handwritten Chinese character recognition engine can realize the self-adaption to the handwriting style of the specific user, and the recognition accuracy is improved.

Description

A kind of writer's adaptive approach based on weight increment secondary Discrimination Functions
Technical field
The invention belongs to the technical field of utilizing computer-processing equipment identification handwriting image, particularly relate to a kind of writer's adaptive approach based on weight increment secondary Discrimination Functions.
Technical background
Handwritten Kanji recognition is meant that generally the user passes through handwriting input device (such as handwriting pad, touch-screen, mouse etc.) writing Chinese characters, and the Chinese-character writing track that simultaneous computer collects handwriting input device is converted to the recognition technology of corresponding Chinese character machine inner code.The common input mode that adopts of traditional handwriting recognition technology is monocase identification, promptly writes Chinese character of a Chinese Character Recognition.Recognition engine of using and user are irrelevant, and promptly recognition engine trains out by a large amount of training samples in advance, and for different user, model and parameter that recognition engine is used all are the same, are trained and are set by the developer in advance.Owing to adopted the training sample of big data quantity, so recognition engine can satisfy the user's of normalized written recognition accuracy requirement.But the writing style of different user is widely different, each user writing style comprises own personalized place toward contact except having general character, the recognition engine that has nothing to do with the user is during at this class user during writing Chinese characters, recognition accuracy is often unsatisfactory, haves much room for improvement.
Summary of the invention
The objective of the invention is to overcome the deficiency that traditional recognition engine can't adapt to specific user's writing style, thereby provide a kind of recognition engine that allows dynamically to adapt to the method that specific user's writing style improves discrimination.
The technical solution used in the present invention is:
A kind of writer's adaptive approach based on weight increment secondary Discrimination Functions, its step is as follows:
(1), chooses a spot of increment sample of specific user;
(2), to increment sample extraction feature, and the increment sample characteristics is carried out linear discriminant analysis (LDA) conversion according to original linear discriminant analysis (LDA) model;
(3), utilize the increment sample and, dynamically update the mean vector and the covariance matrix based on weighting of each classification in conjunction with increment secondary Discrimination Functions (WIMQDF) algorithm of weighting;
(4), the mean vector and the covariance matrix based on weighting of each classification after adopt upgrading, the secondary Discrimination Functions sorter of retrofit.
Described step (1) is used for more new template and recognition engine for selecting a spot of sample of specific user.The selection of user's sample should be able to demonstrate fully user's writing style.If original sample is X ={ x i (i=1 ..., N), NBe sample number, and establish its classification number and be MIf the increment sample is Y ={ y j (j=1 ..., L), LBe the increment sample number, and establish its classification number and be PTotal sample after then merging can be expressed as Z = X ∪ Y ={ z k (k=1 ..., L+N), total sample number is L+N, classification adds up to C, and C 〉=M, C 〉=PBe without loss of generality, our hypothesis is for the in the total sample after merging I (i=1 ..., C)Class is respectively at original sample NWith the increment sample LIn have n i With l i Individual sample.Therefore, for merging the new sample in back, belong to the I (i=1 ..., C)The sample number of class is s i =n i + l i
Described step (2) is to line linearity discriminatory analysis (LDA) conversion of going forward side by side of increment sample extraction feature, and its step comprises:
(A), the increment sample characteristics extracts, and for each increment sample, all adopts all directions to extract its direction character to feature extracting method;
(B), the increment sample characteristics is carried out linear discriminant analysis (LDA) conversion, the purpose of LDA conversion is that each Chinese character classification can farthest be separated, and establishes through from all directions after feature extraction classification thereby improve discrimination I (i=1 ..., C)Original feature vector be
Figure 938747DEST_PATH_IMAGE001
, make original linear discriminant analysis transformation matrix be
Figure 518764DEST_PATH_IMAGE002
, establish through the proper vector after the linear discriminant analysis conversion and be
Figure 60603DEST_PATH_IMAGE003
, then try to achieve by following formula
Figure 684352DEST_PATH_IMAGE003
:
Figure 718167DEST_PATH_IMAGE004
。
Described step (3) is utilized the increment sample and in conjunction with increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, is dynamically updated the mean vector and the covariance matrix based on weighting of each classification, and its step comprises:
(A), establish classification I (i=1 ..., C)Through the original mean vector after the linear discriminant analysis conversion be
Figure 960929DEST_PATH_IMAGE005
, the original sample number is
Figure 484839DEST_PATH_IMAGE006
, then try to achieve by following formula
Figure 154855DEST_PATH_IMAGE005
:
Figure 675966DEST_PATH_IMAGE007
(B), establish classification I (i=1 ..., C)Through the increment sample average vector after the linear discriminant analysis conversion be , the increment number of samples is
Figure 97906DEST_PATH_IMAGE009
, according to increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, classification I (i=1 ..., C)Through the increment sample after the linear discriminant analysis conversion based on the mean vector of weighting with
Figure 814189DEST_PATH_IMAGE008
Unanimity is then tried to achieve by following formula
Figure 416072DEST_PATH_IMAGE008
:
Figure 125271DEST_PATH_IMAGE010
(C), according to the result of (A), (B), adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then establish and merge the mean vector based on weighting of the new sample in back after and be through the linear discriminant analysis conversion , try to achieve by following formula:
(D), ask classification in the original sample I (i=1 ..., C)Covariance matrix, classification in the original sample iThrough the mean vector after the linear discriminant analysis conversion be
Figure 206993DEST_PATH_IMAGE005
, number of samples is
Figure 251041DEST_PATH_IMAGE006
, its covariance matrix then
Figure 351853DEST_PATH_IMAGE013
Try to achieve by following formula:
Figure 534572DEST_PATH_IMAGE014
(E), ask classification in the increment sample I (i=1 ..., C)Covariance matrix, classification in the increment sample iThrough the mean vector after the linear discriminant analysis conversion be
Figure 244470DEST_PATH_IMAGE008
, number of samples is
Figure 170838DEST_PATH_IMAGE009
, adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then based on the covariance matrix of weighting
Figure 126155DEST_PATH_IMAGE015
Try to achieve by following formula:
Figure 401148DEST_PATH_IMAGE016
(F), ask classification in the new sample in merging back I (i=1 ..., C)Covariance matrix, according to the result of (D), (E), adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then merge the covariance matrix of sample based on weighting For:
。
Described step (4) adopts the mean vector and the covariance matrix of each classification after upgrading, the secondary Discrimination Functions sorter of retrofit, and its step comprises:
(A), establish
Figure 395014DEST_PATH_IMAGE019
Represent iIndividual classification (i=1 ..., C), With
Figure 345970DEST_PATH_IMAGE021
Represent the merging sample that obtains according to step (3) mean vector and covariance matrix respectively, and the prior probability of establishing each classification equates that then original secondary Discrimination Functions (QDF) is tried to achieve by following formula based on weighting:
Figure 269933DEST_PATH_IMAGE022
(B), according to Karhunen-Loeve transformation, to covariance matrix
Figure 324476DEST_PATH_IMAGE021
Carry out diagonalization, try to achieve:
Figure 160845DEST_PATH_IMAGE023
Wherein, Λ i = Diag[ λ I1 ..., λ ID ], λ Ij , J=1 ..., DIt is covariance matrix Eigenwert, DIt is the dimension of feature.Φ i =[ φ I1 ..., φ ID ], φ Ij , J=1 ..., DIt is the characteristic of correspondence vector.Φ iBe orthonormal, Φ T iΦ i = I .
(C), according to above-mentioned formula, original secondary Discrimination Functions (QDF) is write as the form of proper vector and eigenwert:
Figure 398633DEST_PATH_IMAGE024
(D), use constant
Figure 448629DEST_PATH_IMAGE025
Substitute less eigenwert, establish KTake the number of proper vector as the leading factor, just obtained final improved secondary Discrimination Functions (MQDF) sorter, try to achieve by following formula:
Figure 314953DEST_PATH_IMAGE026
Through (1) ~ (4) steps, upgrade based on writer's adaptive process of weight increment secondary Discrimination Functions (WIMQDF) and to finish.
The present invention has proposed and user-dependent hand-written adaptive technique first, handwriting recognition engine can be adjusted its model of cognition and parameter according to user's writing style automatically automatically, the recognition system that the user is had nothing to do changes user-dependent recognition system into, thereby improved recognition accuracy greatly to associated user's handwritten Chinese character, simultaneously, this recognition engine is for other users, and its recognition accuracy still can remain unchanged substantially.
Description of drawings
Fig. 1 is a system architecture diagram of the present invention;
Fig. 2 is of the present invention to increment sample extraction feature, and the FB(flow block) of the increment sample characteristics being carried out linear discriminant analysis (LDA) conversion according to original linear discriminant analysis (LDA) model;
Fig. 3 of the present inventionly utilizes the increment sample and in conjunction with increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, dynamically updates the FB(flow block) based on the mean vector and the covariance matrix of weighting of each classification;
Fig. 4 is the FB(flow block) that recognition engine is carried out Chinese Character Recognition.
Embodiment
The present invention is described further below in conjunction with accompanying drawing, implement the used identification equipment of the present invention and can adopt the handwriting pad writing Chinese characters, discern with computing machine, with pure flat escope explicit user graphical interfaces, can adopt the C language to work out all kinds of handling procedures, just can implement the present invention preferably.
System architecture of the present invention adopts from all directions and to feature extracting method the increment sample is carried out feature extraction, and carry out linear discriminant analysis (LDA) conversion by original linear discriminant analysis (LDA) model as shown in Figure 1.Utilize the increment sample then and, dynamically update the mean vector and the covariance matrix of each classification in conjunction with increment secondary Discrimination Functions (WIMQDF) algorithm of weighting; At last by new masterplate mean vector and the improved secondary Discrimination Functions of covariance matrix update (MQDF) sorter.
The a spot of sample of selection user among the present invention is used for more new template and recognition engine, and specific implementation method is when user's sample is selected, and should be able to demonstrate fully user's writing style.If original sample is X ={ x i (i=1 ..., N), NBe sample number, and establish its classification number and be MIf the increment sample is Y ={ y j (j=1 ..., L), LBe the increment sample number, and establish its classification number and be PTotal sample after then merging can be expressed as Z = X ∪ Y ={ z k (k=1 ..., L+N), total sample number is L+N, classification adds up to C, and C 〉=M, C 〉=PBe without loss of generality, our hypothesis is for the in the total sample after merging I (i=1 ..., C)Class is respectively at original sample NWith the increment sample LIn have n i With l i Individual sample.Therefore, for merging the new sample in back, belong to the I (i=1 ..., C)The sample number of class is s i =n i + l i
Among the present invention the increment sample is carried out feature extraction, and the increment sample characteristics is carried out linear discriminant analysis (LDA) conversion process as shown in Figure 2 according to original linear discriminant analysis (LDA) model, specifically comprise following two rapid:
(A), the increment sample characteristics extracts, and for each increment sample, all adopts all directions to extract its direction character to feature extracting method,
(B), the increment sample characteristics is carried out linear discriminant analysis (LDA) conversion, the purpose of LDA conversion is that each Chinese character classification can farthest be separated, and establishes through from all directions after feature extraction classification thereby improve discrimination I (i=1 ..., C)Original feature vector be
Figure 699667DEST_PATH_IMAGE001
, make original linear discriminant analysis transformation matrix be
Figure 716165DEST_PATH_IMAGE002
, establish through the proper vector after the linear discriminant analysis conversion and be
Figure 745301DEST_PATH_IMAGE003
, then try to achieve by following formula
Figure 907161DEST_PATH_IMAGE003
:
Figure 654537DEST_PATH_IMAGE004
Among the present invention utilize the increment sample and in conjunction with increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, dynamically update each classification based on the mean vector of weighting and covariance matrix process as shown in Figure 3, specifically comprise following six steps:
(A), establish classification I (i=1 ..., C)Through the original mean vector after the linear discriminant analysis conversion be
Figure 943567DEST_PATH_IMAGE005
, the original sample number is
Figure 827209DEST_PATH_IMAGE006
, then try to achieve by following formula
Figure 425550DEST_PATH_IMAGE005
:
Figure 394643DEST_PATH_IMAGE007
(B), establish classification I (i=1 ..., C)Through the increment sample average vector after the linear discriminant analysis conversion be
Figure 221784DEST_PATH_IMAGE008
, the increment number of samples is
Figure 878375DEST_PATH_IMAGE009
, according to increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, classification I (i=1 ..., C)Through the increment sample after the linear discriminant analysis conversion based on the mean vector of weighting with Unanimity is then tried to achieve by following formula
Figure 589159DEST_PATH_IMAGE008
:
(C), according to the result of (A), (B), adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then establish and merge the mean vector based on weighting of the new sample in back after and be through the linear discriminant analysis conversion
Figure 327494DEST_PATH_IMAGE011
, try to achieve by following formula:
Figure 143003DEST_PATH_IMAGE012
(D), ask classification in the original sample I (i=1 ..., C)Covariance matrix, classification in the original sample iThrough the mean vector after the linear discriminant analysis conversion be , number of samples is
Figure 255633DEST_PATH_IMAGE006
, its covariance matrix then
Figure 93008DEST_PATH_IMAGE013
Try to achieve by following formula:
Figure 79418DEST_PATH_IMAGE014
(E), ask classification in the increment sample I (i=1 ..., C)Covariance matrix, classification in the increment sample iThrough the mean vector after the linear discriminant analysis conversion be
Figure 385766DEST_PATH_IMAGE008
, number of samples is
Figure 483035DEST_PATH_IMAGE009
, adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then based on the covariance matrix of weighting
Figure 174916DEST_PATH_IMAGE015
Try to achieve by following formula:
Figure 738752DEST_PATH_IMAGE016
(F), ask classification in the new sample in merging back I (i=1 ..., C)Covariance matrix, according to the result of (D), (E), adopt increment secondary Discrimination Functions (WIMQDF) algorithm of weighting, and establish weighting coefficient and be r, then merge the covariance matrix of sample based on weighting
Figure 657030DEST_PATH_IMAGE017
For:
Figure 419974DEST_PATH_IMAGE018
。
The mean vector and the covariance matrix of each classification after the employing among the present invention is upgraded, the secondary Discrimination Functions sorter of retrofit, its step comprises:
(A), establish
Figure 107307DEST_PATH_IMAGE019
Represent iIndividual classification (i=1 ..., C), With
Figure 450881DEST_PATH_IMAGE021
Expression merges mean vector and the covariance matrix of sample based on weighting respectively, and the prior probability of establishing each classification equates that then original secondary Discrimination Functions (QDF) is tried to achieve by following formula:
(B), according to Karhunen-Loeve transformation, to covariance matrix
Figure 228530DEST_PATH_IMAGE021
Carry out diagonalization, try to achieve:
Figure 727645DEST_PATH_IMAGE023
Wherein, Λ i = Diag[ λ I1 ..., λ ID ], λ Ij , J=1 ..., DIt is covariance matrix
Figure 10727DEST_PATH_IMAGE021
Eigenwert, DIt is the dimension of feature.Φ i =[ φ I1 ..., φ ID ], φ Ij , J=1 ..., DIt is the characteristic of correspondence vector.Φ iBe orthonormal, Φ T iΦ i = I .
(C), according to above-mentioned formula, original secondary Discrimination Functions (QDF) is write as the form of proper vector and eigenwert:
Figure 863277DEST_PATH_IMAGE024
(D), use constant
Figure 525202DEST_PATH_IMAGE025
Substitute less eigenwert, establish KTake the number of proper vector as the leading factor, just obtained final improved secondary Discrimination Functions (MQDF) sorter, try to achieve by following formula:
Figure 319852DEST_PATH_IMAGE026
。
Adopt the writer's adaptive technique based on weight increment secondary Discrimination Functions (WIMQDF) of the present invention to obtain confirmation with the experiment of its excellent performance by large sample and big classification.Utilize flow process that recognition engine of the present invention carries out Chinese Character Recognition as shown in Figure 4.
Describe below and adopt method of the present invention, to the result who carries out related experiment of a large amount of on-line handwritten Chinese character samples.
The present invention adopts by the WORD8888 Chinese character phrase sample data subclass in the SCUT-COUCH hand script Chinese input equipment sample database of this laboratory collection and open issue as experimental data, this subclass comprises 130 Writer's 8888 Chinese characters in common use phrase samples of difference, and every writer independently writes and finishes a cover sample data.We select wherein 30 covers as experimental data, and manually these phrases are cut out entirely and become individual character.Individual character in these phrases is Chinese characters of level 2, through statistical study, after these phrases are cut into individual character, comprises 2078 class first-level Chinese characters altogether, totally 19595 Chinese characters.We become database A to the individual character character library that this 30 cover is cut out by phrase.
For database A, the number of samples of each class, 50% of picked at random sum carries out the training of increment MQDF model, and remaining 50% carries out the test of user writing style adaptive performance.
Following table is listed is based on different weighting coefficients r, 30 sets of data do not adopt the present invention and the contrast of adopting the present invention to the average recognition rate of writing Chinese characters among the database A.
Figure DEST_PATH_IMAGE027
Can see that behind employing the present invention, discrimination is greatly improved.Wherein, when rGot 0.3 o'clock, discrimination reaches mxm..
What following table was listed is, when rGot 0.3 o'clock, and, do not adopt the present invention and the discrimination that adopts the present invention to writing Chinese characters for every suit data among the database A (being a certain specific user).
The user Do not adopt adaptive technique Adopt adaptive technique The mistake rate of descent The user Do not adopt adaptive technique Adopt adaptive technique The mistake rate of descent
1 92.84% 96.48% 50.74% 16 94.64% 98.10% 64.52%
2 86.44% 95.10% 63.88% 17 79.96% 91.66% 58.41%
3 85.83% 96.38% 74.44% 18 86.31% 92.90% 48.14%
4 95.60% 98.61% 68.34% 19 61.29% 84.31% 59.45%
5 85.38% 95.44% 68.79% 20 59.78% 80.18% 50.72%
6 96.39% 98.75% 65.42% 21 72.58% 88.97% 59.79%
7 83.99% 94.72% 67.01% 22 92.14% 97.73% 71.15%
8 82.07% 93.66% 64.65% 23 90.90% 96.91% 66.03%
9 86.14% 95.05% 64.31% 24 53.95% 85.08% 67.61%
10 85.33% 94.99% 65.82% 25 93.86% 97.45% 58.37%
11 93.63% 97.55% 61.54% 26 88.86% 95.93% 63.45%
12 92.99% 97.21% 60.14% 27 93.04% 97.26% 60.69%
13 95.51% 98.50% 66.67% 28 91.01% 95.63% 51.39%
14 92.66% 98.18% 75.26% 29 81.77% 94.42% 69.44%
15 92.65% 97.63% 67.71% 30 94.54% 97.27% 50.09%
On average 86.07% 94.74% 62.21% ? ? ? ?
As seen from table, for the user of appointment, adopt writer's adaptive technique based on weight increment secondary Discrimination Functions (WIMQDF) after, this user's Chinese Character Recognition rate is had by a relatively large margin raising.

Claims (5)

1. the writer's adaptive approach based on weight increment secondary Discrimination Functions is characterized in that comprising the steps:
(1), chooses and be used for the more specific user's of new template and recognition engine increment sample;
(2), to increment sample extraction feature, and the increment sample characteristics is carried out the linear discriminant analysis conversion according to original linear discriminant analysis model;
(3), utilize the increment sample and, dynamically update the mean vector and the covariance matrix based on weighting of each classification in conjunction with the increment secondary Discrimination Functions algorithm of weighting;
(4), the mean vector and the covariance matrix based on weighting of each classification after adopt upgrading, the secondary Discrimination Functions sorter of retrofit.
2. the writer's adaptive approach based on weight increment secondary Discrimination Functions according to claim 1 is characterized in that the concrete operations of described step (1) are as follows:
The setting original sample is X ={ x i (i=1 ..., N), NBe sample number, and establish its classification number and be MIf the increment sample is Y ={ y j (j=1 ..., L), LBe the increment sample number, and establish its classification number and be P, the total schedule of samples after then merging is shown Z = X ∪ Y ={ z k (k=1 ..., L+N), total sample number is L+N, classification adds up to C, and C 〉=M, C 〉=P, for the in the total sample after merging I (i=1 ..., C)Class is respectively at original sample NWith the increment sample LIn have n i With l i Individual sample merges the new sample in back, belongs to the I (i=1 ..., C)The sample number of class is s i =n i + l i
3. the writer's adaptive approach based on weight increment secondary Discrimination Functions according to claim 2 is characterized in that the concrete step of updating of described step (2) is as follows:
(21), the increment sample characteristics extracts, and for each increment sample, all adopts all directions to extract its direction character to feature extracting method,
(22), the increment sample characteristics is carried out the linear discriminant analysis conversion, establish through from all directions after feature extraction classification I (i=1 ..., C)Original feature vector be
Figure 348376DEST_PATH_IMAGE001
, make original linear discriminant analysis transformation matrix be
Figure 421374DEST_PATH_IMAGE002
, establish through the proper vector after the linear discriminant analysis conversion and be
Figure 920489DEST_PATH_IMAGE003
, then try to achieve by formula
Figure 547779DEST_PATH_IMAGE003
:
Figure 790542DEST_PATH_IMAGE004
4. the writer's adaptive approach based on weight increment secondary Discrimination Functions according to claim 3 is characterized in that the concrete step of updating of described step (3) is as follows:
(31), establish classification I (i=1 ..., C)Through the original mean vector after the linear discriminant analysis conversion be
Figure 452467DEST_PATH_IMAGE005
, the original sample number is
Figure 122483DEST_PATH_IMAGE006
, then try to achieve by following formula
Figure 980280DEST_PATH_IMAGE005
:
Figure 26733DEST_PATH_IMAGE007
(32), establish classification I (i=1 ..., C)Through the increment sample average vector after the linear discriminant analysis conversion be , the increment number of samples is
Figure 384082DEST_PATH_IMAGE009
, according to the increment secondary Discrimination Functions algorithm of weighting, classification I (i=1 ..., C)Through the increment sample after the linear discriminant analysis conversion based on the mean vector of weighting with
Figure 720385DEST_PATH_IMAGE008
Unanimity is then tried to achieve by following formula :
Figure 675889DEST_PATH_IMAGE010
(33), according to the result of step (31), (32), adopt the increment secondary Discrimination Functions algorithm of weighting, and establish weighting coefficient and be r, then establish and merge the mean vector based on weighting of the new sample in back after and be through the linear discriminant analysis conversion
Figure 687707DEST_PATH_IMAGE011
, try to achieve by following formula:
Figure 776886DEST_PATH_IMAGE012
(34), calculate classification in the original sample I (i=1 ..., C)Covariance matrix, classification in the original sample iThrough the mean vector after the linear discriminant analysis conversion be , number of samples is
Figure 921745DEST_PATH_IMAGE006
, its covariance matrix then Try to achieve by following formula:
(35), calculate classification in the increment sample I (i=1 ..., C)Covariance matrix, classification in the increment sample iThrough the mean vector after the linear discriminant analysis conversion be
Figure 341728DEST_PATH_IMAGE008
, number of samples is
Figure 687259DEST_PATH_IMAGE009
, adopt the increment secondary Discrimination Functions algorithm of weighting, and establish weighting coefficient and be r, then based on the covariance matrix of weighting
Figure 43810DEST_PATH_IMAGE015
Try to achieve by following formula:
Figure 842001DEST_PATH_IMAGE016
(36), calculate classification in the new sample in merging back I (i=1 ..., C)Covariance matrix, according to the result of step (34), (35), adopt the increment secondary Discrimination Functions algorithm of weighting, and establish weighting coefficient and be r, then merge the covariance matrix of sample based on weighting
Figure 572060DEST_PATH_IMAGE017
For:
Figure 772097DEST_PATH_IMAGE018
。
5. the writer's adaptive approach based on weight increment secondary Discrimination Functions according to claim 4 is characterized in that the concrete step of updating of described step (4) is as follows:
(41), set
Figure 296620DEST_PATH_IMAGE019
Represent iIndividual classification (i=1 ..., C), With
Figure 647015DEST_PATH_IMAGE021
Represent the merging sample that obtains according to step (3) mean vector and covariance matrix respectively, and the prior probability of establishing each classification equates that then original secondary Discrimination Functions is tried to achieve by following formula based on weighting:
Figure 170401DEST_PATH_IMAGE022
(42), according to Karhunen-Loeve transformation, to covariance matrix
Figure 865824DEST_PATH_IMAGE021
Carry out diagonalization, try to achieve , wherein, Λ i = Diag[ λ I1 ..., λ ID ], λ Ij , J=1 ..., DIt is covariance matrix
Figure 710469DEST_PATH_IMAGE021
Eigenwert, DBe the dimension of feature, Φ i =[ φ I1 ..., φ ID ], φ Ij , J=1 ..., DBe the characteristic of correspondence vector, Φ iBe orthonormal, Φ T iΦ i = I
(43), according to above-mentioned formula, original secondary Discrimination Functions is write as the form of proper vector and eigenwert:
(44), use constant
Figure 751423DEST_PATH_IMAGE025
Substitute less eigenwert, establish KTake the number of proper vector as the leading factor, just obtained final improved secondary Discrimination Functions sorter, try to achieve by following formula:
Figure 745924DEST_PATH_IMAGE026
。
CN2011101087757A 2011-04-28 2011-04-28 Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF) Pending CN102298704A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2011101087757A CN102298704A (en) 2011-04-28 2011-04-28 Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2011101087757A CN102298704A (en) 2011-04-28 2011-04-28 Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF)

Publications (1)

Publication Number Publication Date
CN102298704A true CN102298704A (en) 2011-12-28

Family

ID=45359113

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2011101087757A Pending CN102298704A (en) 2011-04-28 2011-04-28 Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF)

Country Status (1)

Country Link
CN (1) CN102298704A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103258217A (en) * 2013-05-15 2013-08-21 中国科学院自动化研究所 Pedestrian detection method based on incremental learning
CN103324929A (en) * 2013-06-25 2013-09-25 天津师范大学 Handwritten Chinese character recognition method based on substructure learning

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1652138A (en) * 2005-02-08 2005-08-10 华南理工大学 Method for identifying hand-writing characters
CN101630368A (en) * 2009-08-25 2010-01-20 华南理工大学 Self-adaptive method of user writing style for recognizing handwritten Chinese characters

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1652138A (en) * 2005-02-08 2005-08-10 华南理工大学 Method for identifying hand-writing characters
CN101630368A (en) * 2009-08-25 2010-01-20 华南理工大学 Self-adaptive method of user writing style for recognizing handwritten Chinese characters

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
金连文: "手写体汉字识别的研究", 《中国优秀博硕士学位论文全文数据库（博士） 信息科技辑》, 15 November 2006 (2006-11-15), pages 51 - 70 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103258217A (en) * 2013-05-15 2013-08-21 中国科学院自动化研究所 Pedestrian detection method based on incremental learning
CN103324929A (en) * 2013-06-25 2013-09-25 天津师范大学 Handwritten Chinese character recognition method based on substructure learning
CN103324929B (en) * 2013-06-25 2016-05-18 天津师范大学 Based on the handwritten Chinese recognition methods of minor structure study

Similar Documents

Publication Publication Date Title
CN102622610B (en) Handwritten Uyghur character recognition method based on classifier integration
CN101866417B (en) Method for identifying handwritten Uigur characters
US20070005537A1 (en) Handwriting recognition using a comparative neural network
US20060274943A1 (en) Personalized implicit and explicit character shape adaptation and recognition
CN101630362B (en) Chinese character writing quality evaluation method based on confidence level
CN101976354A (en) Method and device for judging standardization of writing Chinese characters
CN104834941A (en) Offline handwriting recognition method of sparse autoencoder based on computer input
CN101630368B (en) Self-adaptive method of user writing style for recognizing handwritten Chinese characters
Mehrotra et al. Unconstrained handwritten Devanagari character recognition using convolutional neural networks
JP2009506464A (en) Use with handwriting input style
CN107861947A (en) A kind of method of the card language name Entity recognition based on across language resource
CN103020167B (en) A kind of computer Chinese file classification method
Xu et al. Air-writing characters modelling and recognition on modified CHMM
CN105868796A (en) Design method for linear discrimination of sparse representation classifier based on nuclear space
CN108369637A (en) System and method for beautifying digital ink
CN107909003B (en) gesture recognition method for large vocabulary
CN110032948B (en) Sketch gesture recognition method based on interaction time sequence information
CN102184426B (en) Method for enabling handwritten Chinese character recognition engine to dynamically adapt to writing style of user
CN102298704A (en) Writer self-adaptation method based on weighing increment modified quadratic discriminant function (WIMQDF)
CN101604392B (en) Chinese character stroke feature parallel extraction method used for on-line handwritten Chinese character recognition
CN101604393B (en) Chinese character stroke feature extraction method used for on-line handwritten Chinese character recognition
CN107357782A (en) One kind identification user's property method for distinguishing and terminal
CN101697200A (en) Handwritten Chinese grass-style phrase identification method irrelevant to rotation
CN101604378B (en) Chinese character stroke feature serial extraction method used for on-line handwritten Chinese character recognition
CN102184424B (en) Adaptive method increment quadratic discriminant function based on weighted authentication subspace

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20111228