CN110909757A - Method for selecting and updating template of biological recognition system - Google Patents

Method for selecting and updating template of biological recognition system Download PDF

Info

Publication number
CN110909757A
CN110909757A CN201910767514.2A CN201910767514A CN110909757A CN 110909757 A CN110909757 A CN 110909757A CN 201910767514 A CN201910767514 A CN 201910767514A CN 110909757 A CN110909757 A CN 110909757A
Authority
CN
China
Prior art keywords
template
sample
candidate
selection
templates
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.)
Granted
Application number
CN201910767514.2A
Other languages
Chinese (zh)
Other versions
CN110909757B (en
Inventor
岳峰
李彬
陈曦
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.)
BEIJING WHOIS TECHNOLOGY Co Ltd
Original Assignee
BEIJING WHOIS TECHNOLOGY Co Ltd
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 BEIJING WHOIS TECHNOLOGY Co Ltd filed Critical BEIJING WHOIS TECHNOLOGY Co Ltd
Priority to CN201910767514.2A priority Critical patent/CN110909757B/en
Publication of CN110909757A publication Critical patent/CN110909757A/en
Application granted granted Critical
Publication of CN110909757B publication Critical patent/CN110909757B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a method for selecting and updating a biological recognition system template, which relates to digital image processing in biological recognition, and comprises two stages of candidate template selection and sample passing table updating process based on sample importance evaluation and biological recognition system template selection process based on a nearest neighbor classifier, wherein the candidate template selection and sample passing table updating process based on sample importance evaluation comprises the establishment of a sample passing table AsSelecting a candidate template and updating a sample passing table, wherein the selection process of the biological recognition system template based on the nearest neighbor classifier comprises the step of selecting a new template set T according to the joint passing table*And a template candidate space TMThe method overcomes the defects that the prior method for selecting and updating the biological identification system template has no universality and the false rejection rate of the biological identification system is higher due to the change in the template class.

Description

Method for selecting and updating template of biological recognition system
Technical Field
The technical scheme of the invention relates to digital image processing in biological recognition, in particular to a method for selecting and updating a biological recognition system template.
Background
The operation of the existing biometric systems comprises two phases: a characteristic registration stage and an authentication stage, wherein in the characteristic registration stage, an input sample is acquired and preprocessed to finally obtain biological characteristics with difference, and the biological characteristics are stored in a database; in the authentication stage, after similar processing is performed on the input sample, the obtained biological characteristics are used for matching the template stored in the database, and finally the authenticated identity information is obtained.
The biological identification prototype system uses biological characteristics acquired from registrant input as a template, and identity authentication is carried out by using a nearest neighbor classifier after registration is successful. The nearest neighbor classifier is a general name of a class of classifiers which are most widely applied in the field of current biological identification, the most similar template is searched as the output to be simulated by comparing the input sample with the templates stored in the database one by one, and then whether the input sample and the template belong to the same registered user is finally judged according to a preset authentication threshold value. Existing nearest neighbor classifier based recognition systems can only pass samples that satisfy a certain degree of similarity, and some real users are mistakenly rejected due to the existence of intra-class differences. The intra-class difference refers to samples belonging to the same identity, and due to the acquisition method, environmental change and self-change, the differences among individuals are caused, such as different influences of sample imaging under different illumination, the influence of wearing glasses in face recognition, the influence of palm scars in palm print recognition, the influence of finger peeling in fingerprint recognition and the slow change of the face, the palm print and the iris along with the change of time. To solve this problem, existing biometric systems generally store multiple templates, so that the template set has sufficient expressiveness. But considering memory space and matching real-time, the template set can only store a certain amount of registration features. Today, how to select a smaller and more representative template integration is a very challenging topic.
In addition, biometric identification systems face other challenges such as slow or abrupt changes in biometric characteristics, such as scarring or slow changes in texture over time. A robust biometric system should be able to adaptively handle these intra-class changes through biometric system template update strategies. To date, the current biometric system template update framework can be roughly divided into two categories: sequential processing frameworks and batch frameworks, the former being more prevalent. For the sequential processing framework, CN105631443B discloses an updating method and a terminal device for a fingerprint template, which uses a method for determining dead pixels in a texture to update and complete a fingerprint template. CN102708360A discloses a method for generating and automatically updating a fingerprint template, which is a method for constructing a fused template by fusing information of a plurality of fingerprint samples, but this method is only widely used for fingerprint identification, and the samples are assumed to be based on key point alignment, which is not applicable to a non-contact biometric identification system, because the non-contact biometric identification system has a lot of non-linear transformations and is difficult to be based on key point alignment. For the batch processing framework, CN109325327A discloses a process for updating a template used in face recognition, which is a method for judging whether to update a sample to a template by using two similarity thresholds and confidence intervals, and the method simply uses the matching relationship between a single input sample and the template to judge whether to be suitable as the template, does not consider the difference of intra-class oriented changes, and does not process outliers or false samples, thereby easily causing template pollution. In addition, there is also a method for performing cross-update using multi-modal recognition technology, for example, CN101483652A discloses a biometric feature recognition system, which uses dual-modal recognition technology, and this method uses multiple different biometric features, combines the authentication results and confidence degrees of multiple biometric features to perform identity determination and template update, but this method is not suitable for the commonly used single-modal system.
The prior art methods are all identification updating methods aiming at certain or a plurality of specific biological characteristic information, and are not universal biological characteristic identification updating methods, so that the defects that the universality is not available and the false rejection rate of a biological identification system is high due to the change in templates exist.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method comprises two stages of candidate template selection based on sample importance evaluation, a sample passing table updating process and a biological identification system template selection process based on a nearest neighbor classifier, and overcomes the defects that the existing biological identification system template selection and updating method has no universality and the false rejection rate of the biological identification system is high due to the change in a template class.
The technical scheme adopted by the invention for solving the technical problem is as follows: the method for selecting and updating the biological recognition system template comprises two stages of candidate template selection based on sample importance evaluation, a sample passing table updating process and a biological recognition system template selection process based on a nearest neighbor classifier, and specifically comprises the following steps:
the first step, selecting a candidate template based on sample importance evaluation and updating a sample passing table:
(1.1) establishing samples through Table As：
Sample passage Table AsIs a two-dimensional table for recording the passing or failing information of each sample, and the table is used for recording the positive selection template set T and the candidate template set T of the current biological identification systemCWhen a sample of the registered user passes identity authentication, the matching score of the corresponding template in the column of the table is larger than an authentication threshold value, a corresponding element is set to be 1 to indicate that the sample passes authentication, otherwise, the corresponding element is set to be 0, in an initial state, the two-dimensional table is 0 rows K + N columns, the rows correspond to the number of samples of the registered user acquired in an updating round, the columns correspond to the number of the templates, K represents the number of the elements of the positively selected template set, and N represents the number of the elements in the candidate template set, so that the sample passing table A is establisheds；
(1.2) candidate template selection:
the candidate templates are templates containing change information in a certain class and stored before reaching a new update turn, and when the update turn is reached, the candidate template set TCSelecting proper subset as new template set from the candidate space and the positive template set T as total candidate space, applying for authentication by sample S, and when the matching value of the sample S and all templates in the template set is greater than the authentication threshold TsimAnd is less than the update threshold tcloseThen the sample is putThe method is selected as a candidate template, the authentication threshold value represents the judgment of the system on the real identity of the registered user, the matching score of the input sample and the template is greater than the authentication threshold value, the fact that the current user is identified as a real registered user by the system is represented, the updating threshold value is a threshold value which is stricter than the authentication threshold value, namely, a numerical value is higher, the updating threshold value is used for controlling the input sample and the current template not to be too similar, the combination of the two threshold values of the updating threshold value and the authentication threshold value is used for ensuring that the sample belongs to the real user and has certain difference with the template in, when the condition is satisfied, the sample is allowed to participate in the competition of the selection of the subsequent new template set, which indicates that the sample belongs to the real registered user and has a certain difference with the current template, when the sample S applying for authentication satisfies the candidate template selection condition, the sample passes through the sample passing table A established in the step (1.1).sAdding a column at the tail part of the column to indicate that a new candidate template is added to participate in the updating of a subsequent sample passing table, and when the number of the candidate templates reaches a preset value N of the biological identification system, establishing the sample passing table A according to the step (1.1)sTaking the sum of each row as a weight, and deleting the template with the minimum score after sorting until the candidate template is selected;
(1.3) sample update by table:
the sample created in the above step (1.1) is passed through Table AsInputting the sample S in the step (1.2) for applying for authentication, matching all templates in the template set and the candidate template set one by one, and when the matching score between the input sample S and the template t in the template set is larger than the authentication threshold tsimPut the sample through Table AsCorresponding position A ofs(s, t) ═ 1, otherwise AsWhen the input sample S in the step (1.2) passes the application authentication, the sample established in the step (1.1) passes the table AsAdding a row at the tail of the row, and performing the same operation on each input sample S in the updating process, wherein the sample exists in the working period of the whole authentication biological identification system through a table updating stage;
secondly, selecting a template based on the nearest neighbor classifier:
the template selection based on the nearest neighbor classifier is to select the positive template set T and the candidate template set TcUsing the sum as a template candidate space TMIn the template candidate space TMThe most expressive subset is selected as a new positive selection template set T*The method comprises the following steps:
(2.1) selecting a new template set T according to the joint passing table*：
Build template pass Table AtThe table is a sample pass through table a similar to step (1.1) abovesRepresenting a passing relationship between the templates for expressing the template candidate space TMI.e. the current positive selection template set T and the candidate template set TcThe mutual authentication relationship and the importance degree between the two, the template is established and passes through the table AtThe method comprises the following steps: template passage Table AtA two-dimensional table of K + N rows and K + N columns, both "rows" and "columns" corresponding to the template candidate space TMK represents the number of elements in the positive selection template set, N represents the number of elements in the candidate template set, and the template is established through the table AtFor the template candidate space TMTemplate t in (1)MIt is then compared with the template candidate space TMAll the templates are matched one by one when the template t is matchedMAnd TMMiddle and other templates t'MIs greater than the authentication threshold tsimPut the template through the table AtCorresponding position A oft(tM,t'M) 1, otherwise At(tM,t'M) Pass table a on template 0tThe same operation is performed on each input template in the establishing process of (A), and the template passing table A is completedtEstablishing;
passing the newly created template through Table AtThe samples attached to step (1.1) above in a row-wise fashion are passed through Table AsAfter that, a union is established through Table AjIn combination through Table AjRepresents the template candidate space T in the current update periodMBy the importance of each sample in the combination according to Table AjThe permutation and combination relation of the K + N columns is calculated, and the logical OR relation sample of the K + N columns is obtainedSelecting the combination with the highest passing rate as a new template set T*The specific operation method comprises the following steps:
for the positive selection template set T ═ T i1, 2.. K, and a preset authentication threshold tsimThe authenticated sample set χ (T) in the biometric system is expressed by the following formula (1),
Figure BDA0002172433530000031
in the formula (1), Score (t)iS) represents the sample S and the ith template T in the positive selection template set TiWith a matching score between, using the candidate template space TMRepresenting a union set containing K positive selection templates and N candidate templates, and selecting a new positive selection template set T*Is expressed as a process of maximizing the passing rate, as shown in the following equation (2),
Figure BDA0002172433530000041
further, this process is considered to be in the template candidate space TMA process of maximizing the sample passage rate under the influence of (1), i.e., rewriting the formula (2) to the following formula (3),
Figure BDA0002172433530000042
logical OR operation T according to equation (3)M∩ x (T), selecting new combinations of different templates, and finding the combination with the largest number as the new template set T*Where | is the number of elements in the set, the sample from step (1.1) above is passed through Table AsThe same pattern established to establish the template through Table AtSelecting a new template set T*The process of (2) reduces the false rejection rate of the biometric identification system;
(2.2) template candidate space TMRemoval of intermediate outliers:
template candidate space TMThere may be outliers, i.e. the presence of false samples,increasing the false recognition rate of the biometric identification system when outliers are selected from the template set, using the sample of step (1.1) above to pass through Table AsThe template obtained in the above step (2.1) is shown in Table AtMerging or joining of columns through Table AjThen passes a predetermined minimum threshold tpassOutliers pass through Table A according to associationsjThe numerical value of the sum of each column in the template selection process is detected and is excluded before the selection process of a new positive selection template;
and then, finishing the selection and the updating of the biological recognition system template.
In the method for selecting and updating the template of the biological recognition system, the sample is the feature information of unknown identity input into the biological recognition system, the template is named as the biological feature template of a registrant in the biological recognition system and is the feature information of known identity stored in the biological recognition system, all characters 'K' represent the number of elements in the selected template set, and all characters 'N' represent the number of elements in the candidate template set.
The invention has the beneficial effects that: compared with the prior art, the invention has the following prominent substantive characteristics and remarkable progress:
(1) a robust biometric system should be able to adaptively handle intra-class changes of the biometric system through template update strategies. Template update techniques are methods by which a biometric recognition system learns intra-class changes and continuously updates existing templates based on newly entered samples. For a real enrollee, the biometric system stores several characteristics of the person as templates, and in the face of a string of unlabeled newly entered samples, the method combines the existing templates with the newly entered samples to select a suitable subset as a new template set with better representativeness. The invention provides a method for selecting and updating a biological recognition system template, which comprises two stages of an updating process based on sample importance evaluation, a candidate template selecting process and a biological recognition system template selecting process based on a nearest neighbor classifier, and overcomes the defects that the existing biological recognition system template selecting and updating method has no universality and the false rejection rate of a biological recognition system is high due to the change in a template class.
(2) The method can adaptively update the template set of the biological identification system, because of the control of the authentication threshold and the stricter update threshold, the biological identification system can not generate additional false identification phenomenon, meanwhile, according to the authentication thresholds which are different from 0.65 to 0.85, the method can reduce the false rejection rate by about 8 to 35 percent, and along with the increase of the work period of the biological identification system, more intra-class changes are generated, the effect of the method is more obvious, the false rejection rate of the system is further reduced, the generation of the false rejection condition of the biological identification system caused by the intra-class changes of the template is better resisted, and the requirement of the self-adaptability of the biological identification system is better met.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a sample pass Table A of an embodiment of the present inventionsExamples of (2) are shown.
FIG. 2 is a template pass representation A of an embodiment of the present inventiontExamples of (2) are shown.
FIG. 3 is a joint pass table A of an embodiment of the present inventionjExamples of (2) are shown.
FIG. 4 is a table A of joint passage according to an embodiment of the present inventionjExample graphs of outliers are deleted.
Detailed Description
Examples
The first step, selecting a candidate template based on sample importance evaluation and updating a sample passing table:
(1.1) establishing samples through Table As：
Sample passage Table AsIs a two-dimensional table for recording the passing or failing information of each sample, and the table is used for recording the positive selection template set T and the candidate template set T of the current biological identification systemCThe importance degree of each template in the list, the column of the sample passing table corresponds to the template in the current selected template set and the candidate template set, when one sample of the registered user passes the identity authentication and shows that the matching score of the corresponding template in the column of the list is greater than the authentication threshold value, the corresponding element is set to be 1If not, setting 0, in the initial state, the two-dimensional table is 0 row K + N column, the row corresponds to the number of samples of the registered users collected in an updating turn, the column corresponds to the number of templates, K represents the number of elements of the template set being selected, and N represents the number of elements in the candidate template set, thereby establishing a sample passing table As(ii) a In this embodiment, K is 3, N is 2, and fig. 1 is a sample of this embodiment, which is shown in table asThe figure shows that the positive selection template set comprises 3 templates t1、t2、t3I.e. K is 3, the candidate template set has two templates tc1And tc2I.e. N-2, six samples are taken through table asI.e. the six "rows" shown in the figure: s1、S2、S3、S4、S5、S6The value 1 in the box in the figure indicates that the matching score of the corresponding template in the column of the table is greater than the authentication threshold tsimThe corresponding element is set to 1 to indicate that the sample is authenticated, and the value 0 in the square box in the figure indicates that the matching score of the corresponding template in the column of the table is less than the authentication threshold tsimIf the corresponding element is set to 0, the sample authentication cannot pass;
in the prior art, the template updating usually uses batch processing mode, and for a practical biometric system, there are several factors to be considered, namely the number of input samples, and then the time consumed by updating and the consumption of storage space. In the method of the present invention, the samples are passed through Table AsBy the size of table asThe "rows" correspond to the number of samples and the "columns" correspond to the number of templates, which overcomes the above-mentioned drawbacks of the prior art.
(1.2) candidate template selection:
the candidate templates are templates containing change information in a certain class and stored before reaching a new update turn, and when the update turn is reached, the candidate template set TCAnd the positive selection template set T is taken as the total candidate space TMFrom TMSelecting proper subset as new template setThe application for authentication, when the matching value of the template and all templates in the template set is larger than the authentication threshold value tsimAnd is less than the update threshold tcloseSelecting the sample as a candidate template, wherein the authentication threshold value represents the judgment of the system on the real identity of the registered user, the matching score of the input sample and the template is greater than the authentication threshold value, the fact that the current user is identified as a real registered user by the system is indicated, the updating threshold value is a threshold value which is stricter than the authentication threshold value, namely, has a higher numerical value, and is used for controlling the input sample and the current template not to be too similar, and the combination of the two threshold values of the updating threshold value and the authentication threshold value is used for ensuring that the sample belongs to the real user and has a certain difference with the template in the current template set, when the condition is satisfied, the sample is allowed to participate in the competition of the selection of the subsequent new template set, which indicates that the sample belongs to the real registered user and has a certain difference with the current template, when the sample S applying for authentication satisfies the candidate template selection condition, the sample passes through the sample passing table A established in the step (1.1).sAdding a column at the tail part of the column to indicate that a new candidate template is added to participate in the updating of a subsequent sample passing table, and when the number of the candidate templates reaches a preset value N of the biological identification system, establishing the sample passing table A according to the step (1.1)sTaking the sum of each row as a weight, and deleting the template with the minimum score after sorting until the candidate template is selected;
(1.3) sample update by table:
the sample created in the above step (1.1) is passed through Table AsInputting the sample S in the step (1.2) for applying for authentication, matching all templates in the template set and the candidate template set one by one, and when the matching score between the input sample S and the template t in the template set is larger than the authentication threshold tsimPut the sample through Table AsCorresponding position A ofs(s, t) ═ 1, otherwise AsWhen the input sample S in the step (1.2) passes the application authentication, the sample established in the step (1.1) passes the table AsOne row will be added at the end of the row, and the same operation is performed during the update for each input sample S that is present through the table update phaseThe work cycle of the whole authentication biological identification system;
the algorithm for candidate template selection and sample pass table update is shown in table 1 below:
TABLE 1 Algorithm for candidate template selection and sample update through a table
Figure BDA0002172433530000061
Two samples are involved in the update of the sample passing table, the first sample is a sample which has high similarity with a plurality of templates in the sample set, the other sample cannot be matched with all templates in the template set but is matched with some templates in the candidate set, the first sample represents that the sample is regarded as a real registered sample, but the contained information can be replaced by other current templates because the similarity of the rest current templates is too high, the information is not suitable for being selected as a new template, the second sample represents that the sample is not matched with the templates in the template set but can be matched with the candidate templates in the candidate template set, so that the sample is necessary for evaluating the importance of the candidate templates, and the samples except the samples are added into the candidate template queue or directly ignored;
secondly, selecting a template based on the nearest neighbor classifier:
the template selection based on the nearest neighbor classifier is to select the positive template set T and the candidate template set TcUsing the sum as a template candidate space TMIn the template candidate space TMThe most expressive subset is selected as a new positive selection template set T*The method comprises the following steps:
(2.1) selecting a new template set T according to the joint passing table*：
Build template pass Table AtThe table is a sample pass through table a similar to step (1.1) abovesRepresenting a passing relationship between the templates for expressing the template candidate space TMI.e. the current positive selection template set T and the candidate template set TcThe mutual authentication relationship and the importance degree between the two modules are establishedPassage of the panel table AtThe method comprises the following steps: template passage Table AtA two-dimensional table of K + N rows and K + N columns, both "rows" and "columns" corresponding to the template candidate space TMK represents the number of elements in the positive selection template set, N represents the number of elements in the candidate template set, and the template is established through the table AtFor the template candidate space TMTemplate t in (1)MIt is then compared with the template candidate space TMAll the templates are matched one by one when the template t is matchedMAnd TMMiddle and other templates t'MThe matching score between the two is larger than the preset authentication threshold value tsimPut the template through the table AtCorresponding position A oft(tM,t'M) 1, otherwise At(tM,t'M) Pass table a on template 0tThe same operation is performed on each input template in the establishing process of (A), and the template passing table A is completedtEstablishing;
FIG. 2 shows the template passage A of the present embodimenttThe figure shows that the positive selection template set comprises 3 templates t1、t2、t3I.e. K is 3, the candidate template set has two templates tc1And tc2That is, N is 2, the matching relationship between them, and the diagonal elements, such as the first row and the first column, the second row and the second column, …, and the fifth row and the fifth column, are all 1, which indicates that a template is compared with itself and exceeds the authentication threshold, so the score would be 100;
passing the newly created template through Table AtThe samples attached to step (1.1) above in a row-wise fashion are passed through Table AsAfter that, a union is established through Table AjFIG. 3 is a table A of the association passage of this embodimentjFig. 3 shows the end-to-end arrangement of fig. 1 and 2, i.e. the sample is passed through table asPassage through Table A with templatetComposition of associations through Table A in column alignmentjFor subsequent template selection;
in combination through Table AjRepresents the template candidate space T in the current update periodMBy the importance of each sample in the combination according to Table AjOf medium K + N columnsArranging the combination relation, calculating the passing number of the K + N columns of logic or relation samples, and selecting the combination with the highest passing rate as a new template set T*The specific operation method comprises the following steps:
for the positive selection template set T ═ T i1, 2.. K, and a preset authentication threshold tsimThe authenticated sample set χ (T) in the biometric system is expressed by the following formula (1),
Figure BDA0002172433530000071
in the formula (1), Score (t)iS) represents the sample S and the ith template T in the positive selection template set TiWith a matching score between, using the candidate template space TMRepresenting a union set containing K positive selection templates and N candidate templates, and selecting a new positive selection template set T*Is expressed as a process of maximizing the passing rate, as shown in the following equation (2),
Figure BDA0002172433530000081
further, this process is considered to be in the template candidate space TMA process of maximizing the sample passage rate under the influence of (1), i.e., rewriting the formula (2) to the following formula (3),
Figure BDA0002172433530000082
logical OR operation T according to equation (3)M∩ x (T), selecting new combinations of different templates, and finding the combination with the largest number as the new template set T*Where | is the number of elements in the set, the sample from step (1.1) above is passed through Table AsThe same pattern established to establish the template through Table AtSelecting a new template set T*The process of (2) reduces the false rejection rate of the biometric identification system;
selecting a new template set T*The algorithm of (a) is shown in table 2 below:
TABLE 2 selection of new templatesCollection T*Is calculated by
Figure BDA0002172433530000083
(2.2) template candidate space TMRemoval of intermediate outliers:
template candidate space TMThere may be outliers, i.e., false samples, which increase the false recognition rate of the biometric identification system when the outliers are selected from the template set, and the sample passing table A in the step (1.1) is utilizedsThe template obtained in the above step (2.1) is shown in Table AtMerging or joining of columns through Table AjThen passes a predetermined minimum threshold tsimOutliers pass through Table A according to associationsjThe numerical value of the sum of each column in the template selection process is detected and is excluded before the selection process of a new positive selection template;
FIG. 4 is a table A of the passing union passing of the present embodimentjExample graph for outlier deletion, shown by a statistical threshold tpassWhen the value is set to 2, the template set t is selected in the positive way in FIG. 41Will be identified as outliers and will not participate in the selection of the new positive selection template, passing through the statistical threshold tpassWhen the value is set to 4, the template set t is selected in the positive way in FIG. 42And candidate template set tc1Will be identified as outliers and will not participate in the selection of the new positive selection template, passing through the statistical threshold tpassWhen the setting is 6, the template set t is selected in the positive way in FIG. 43Will be identified as outliers and will not participate in the selection of the new positive selection template, passing through the statistical threshold tpassIf the result is set to 3, the candidate template set t in FIG. 4 isc2Will be identified as outliers and will not participate in the selection of the new positive selection template;
and then, finishing the selection and the updating of the biological recognition system template.
In the above embodiment, the sample is feature information of unknown identity input into the biometric system, the template is a biometric template of a registrant in the biometric system, and is feature information of known identity stored in the biometric system, all characters "K" represent the number of elements in the template set being selected, and some characters "N" represent the number of elements in the candidate template set.

Claims (3)

1. The method for selecting and updating the biological recognition system template is characterized by comprising the following steps: the method comprises two stages of a candidate template selection process based on sample importance evaluation, a sample passing table updating process and a biological recognition system template selection process based on a nearest neighbor classifier, and specifically comprises the following steps:
the first step, selecting a candidate template based on sample importance evaluation and updating a sample passing table:
(1.1) establishing samples through Table As：
Sample passage Table AsIs a two-dimensional table for recording the passing or failing information of each sample, and the table is used for recording the positive selection template set T and the candidate template set T of the current biological identification systemCWhen a sample of the registered user passes identity authentication, the matching score of the corresponding template in the column of the table is larger than an authentication threshold value, a corresponding element is set to be 1 to indicate that the sample passes authentication, otherwise, the corresponding element is set to be 0, in an initial state, the two-dimensional table is 0 rows K + N columns, the rows correspond to the number of samples of the registered user acquired in an updating round, the columns correspond to the number of the templates, K represents the number of the elements of the positively selected template set, and N represents the number of the elements in the candidate template set, so that the sample passing table A is establisheds；
(1.2) candidate template selection:
the candidate templates are templates containing change information in a certain class and stored before reaching a new update turn, and when the update turn is reached, the candidate template set TCSelecting proper subset as new template set from the candidate space and the positive template set T as total candidate space, applying for authentication by sample S, and when the matching value of the sample S and all templates in the template set is greater than the authentication threshold TsimAnd is less than the update threshold tcloseThen, the sample is selected as a candidate template, the authentication threshold value represents the judgment of the system on the real identity of the registered user, and the matching score of the input sample and the template is greater than that of the authentication threshold value tableThe combination of the two thresholds is used for ensuring that the sample belongs to the real registered user and has a certain difference with the template in the current template set, when the condition is met, the sample belongs to the real registered user and has a certain difference with the current template set, the sample is allowed to participate in competition of selection of a subsequent new template set, and when the sample S applying for authentication meets the candidate template selection condition, the sample passes through the sample passing table A established in the step (1.1)sAdding a column at the tail part of the column to indicate that a new candidate template is added to participate in the updating of a subsequent sample passing table, and when the number of the candidate templates reaches a preset value N of the biological identification system, establishing the sample passing table A according to the step (1.1)sTaking the sum of each row as a weight, and deleting the template with the minimum score after sorting until the candidate template is selected;
(1.3) sample update by table:
the sample created in the above step (1.1) is passed through Table AsInputting the sample S in the step (1.2) for applying for authentication, matching all templates in the template set and the candidate template set one by one, and when the matching score between the input sample S and the template t in the template set is larger than the authentication threshold tsimPut the sample through Table AsCorresponding position A ofs(s, t) ═ 1, otherwise AsWhen the input sample S in the step (1.2) passes the application authentication, the sample established in the step (1.1) passes the table AsAdding a row at the tail of the row, and performing the same operation on each input sample S in the updating process, wherein the sample exists in the working period of the whole authentication biological identification system through a table updating stage;
secondly, selecting a template based on the nearest neighbor classifier:
the template selection based on the nearest neighbor classifier is to select the positive template set T and the candidate template set TcUsing the sum as a template candidate space TMIn the template candidate space TMThe most expressive subset is selected as a new positive selection template set T*The method comprises the following steps:
(2.1) selecting a new template set T according to the joint passing table*：
Build template pass Table AtThe table is a sample pass through table a similar to step (1.1) abovesRepresenting a passing relationship between the templates for expressing the template candidate space TMI.e. the current positive selection template set T and the candidate template set TcThe mutual authentication relationship and the importance degree between the two, the template is established and passes through the table AtThe method comprises the following steps: template passage Table AtA two-dimensional table of K + N rows and K + N columns, both "rows" and "columns" corresponding to the template candidate space TMK represents the number of elements in the positive selection template set, N represents the number of elements in the candidate template set, and the template is established through the table AtFor the template candidate space TMTemplate t in (1)MIt is then compared with the template candidate space TMAll the templates are matched one by one when the template t is matchedMAnd TMMiddle and other templates t'MIs greater than the authentication threshold tsimPut the template through the table AtCorresponding position A oft(tM,t'M) 1, otherwise At(tM,t'M) Pass table a on template 0tThe same operation is performed on each input template in the establishing process of (A), and the template passing table A is completedtEstablishing;
passing the newly created template through Table AtThe samples attached to step (1.1) above in a row-wise fashion are passed through Table AsAfter that, a union is established through Table AjIn combination through Table AjRepresents the template candidate space T in the current update periodMBy the importance of each sample in the combination according to Table AjThe permutation and combination relation of the K + N columns is calculated, the passing number of the logic or relation samples of the K + N columns is calculated, and the combination with the highest passing rate is selected as a new template set T*The specific operation method comprises the following steps:
for positive selection template setT＝ti1, 2.. K, and a preset authentication threshold tsimThe authenticated sample set χ (T) in the biometric system is expressed by the following formula (1),
Figure FDA0002172433520000021
in the formula (1), Score (t)iS) represents the sample S and the ith template T in the positive selection template set TiWith a matching score between, using the candidate template space TMRepresenting a union set containing K positive selection templates and N candidate templates, and selecting a new positive selection template set T*Is expressed as a process of maximizing the passing rate, as shown in the following equation (2),
Figure FDA0002172433520000022
further, this process is considered to be in the template candidate space TMA process of maximizing the sample passage rate under the influence of (1), i.e., rewriting the formula (2) to the following formula (3),
Figure FDA0002172433520000023
logical OR operation T according to equation (3)M∩ x (T), selecting new combinations of different templates, and finding the combination with the largest number as the new template set T*Where | is the number of elements in the set, the sample from step (1.1) above is passed through Table AsThe same pattern established to establish the template through Table AtSelecting a new template set T*The process of (2) reduces the false rejection rate of the biometric identification system;
(2.2) template candidate space TMRemoval of intermediate outliers:
template candidate space TMThere may be outliers, i.e., false samples, which increase the false recognition rate of the biometric identification system when the outliers are selected from the template set, and the sample passing table A in the step (1.1) is utilizedsWith the above (2.1)Template passage of StepstMerging or joining of columns through Table AjThen passes a predetermined minimum threshold tpassOutliers pass through Table A according to associationsjThe numerical value of the sum of each column in the template selection process is detected and is excluded before the selection process of a new positive selection template;
and then, finishing the selection and the updating of the biological recognition system template.
2. The biometric system template selection and update method of claim 1, wherein: the algorithm for candidate template selection and sample pass table update is as follows:
Figure FDA0002172433520000031
3. the biometric system template selection and update method of claim 1, wherein: selecting a new template set T*The algorithm of (a) is as follows:
Figure FDA0002172433520000032
CN201910767514.2A 2019-08-20 2019-08-20 Method for selecting and updating template of biological recognition system Active CN110909757B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910767514.2A CN110909757B (en) 2019-08-20 2019-08-20 Method for selecting and updating template of biological recognition system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910767514.2A CN110909757B (en) 2019-08-20 2019-08-20 Method for selecting and updating template of biological recognition system

Publications (2)

Publication Number Publication Date
CN110909757A true CN110909757A (en) 2020-03-24
CN110909757B CN110909757B (en) 2023-07-14

Family

ID=69814444

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910767514.2A Active CN110909757B (en) 2019-08-20 2019-08-20 Method for selecting and updating template of biological recognition system

Country Status (1)

Country Link
CN (1) CN110909757B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112528783A (en) * 2020-11-30 2021-03-19 深圳邦健生物医疗设备股份有限公司 Electrocardiogram heartbeat data clustering method and device, electronic equipment and medium
WO2023083051A1 (en) * 2021-11-15 2023-05-19 北京眼神智能科技有限公司 Biometric recognition method and apparatus, and device and storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070150745A1 (en) * 2005-12-22 2007-06-28 Michael Peirce Biometric authentication system
US20180373717A1 (en) * 2015-12-16 2018-12-27 Thomson Licensing Methods and apparatuses for processing biometric responses to multimedia content
CN109934114A (en) * 2019-02-15 2019-06-25 重庆工商大学 A kind of finger vena template generation and more new algorithm and system
CN109981290A (en) * 2019-03-26 2019-07-05 南京信息工程大学 The communication system and method close based on no certificate label under a kind of intelligent medical environment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070150745A1 (en) * 2005-12-22 2007-06-28 Michael Peirce Biometric authentication system
US20180373717A1 (en) * 2015-12-16 2018-12-27 Thomson Licensing Methods and apparatuses for processing biometric responses to multimedia content
CN109934114A (en) * 2019-02-15 2019-06-25 重庆工商大学 A kind of finger vena template generation and more new algorithm and system
CN109981290A (en) * 2019-03-26 2019-07-05 南京信息工程大学 The communication system and method close based on no certificate label under a kind of intelligent medical environment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
周玲丽;赖剑煌;: "生物特征数据安全保护技术的发展" *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112528783A (en) * 2020-11-30 2021-03-19 深圳邦健生物医疗设备股份有限公司 Electrocardiogram heartbeat data clustering method and device, electronic equipment and medium
CN112528783B (en) * 2020-11-30 2024-04-16 深圳邦健生物医疗设备股份有限公司 Electrocardiogram and heart beat data clustering method, device, electronic equipment and medium
WO2023083051A1 (en) * 2021-11-15 2023-05-19 北京眼神智能科技有限公司 Biometric recognition method and apparatus, and device and storage medium

Also Published As

Publication number Publication date
CN110909757B (en) 2023-07-14

Similar Documents

Publication Publication Date Title
Assegie et al. Handwritten digits recognition with decision tree classification: a machine learning approach
Xiao et al. Learning deep feature representations with domain guided dropout for person re-identification
US11232280B2 (en) Method of extracting features from a fingerprint represented by an input image
CN109934114B (en) Finger vein template generation and updating algorithm and system
CN109002755B (en) Age estimation model construction method and estimation method based on face image
Kumar et al. An improved biometric fusion system of fingerprint and face using whale optimization
CN109145704B (en) Face portrait recognition method based on face attributes
Ghoualmi et al. An efficient feature selection scheme based on genetic algorithm for ear biometrics authentication
CN110909757A (en) Method for selecting and updating template of biological recognition system
TWI783199B (en) Processing method of face recognition and electronic device
Yang et al. A Face Detection Method Based on Skin Color Model and Improved AdaBoost Algorithm.
Ge et al. Deep and discriminative feature learning for fingerprint classification
WO2018137226A1 (en) Fingerprint extraction method and device
Lv et al. Face detection based on skin color and AdaBoost algorithm
Arjun et al. Multimodal biometric recognition: fusion of modified adaptive bilinear interpolation data samples of face and signature using local binary pattern features
Mayya et al. Iris and palmprint decision fusion to enhance human recognition
JP2005165447A (en) Age/sex discrimination device, age/sex discrimination method and person recognition device
Jha et al. Ubsegnet: Unified biometric region of interest segmentation network
CN112116012B (en) Finger vein instant registration and identification method and system based on deep learning
CN105488493B (en) A kind of quick iris identification method
Shukla et al. A Hybrid Model of Multimodal Biometrics System using Fingerprint and Face as Traits
CN111382703A (en) Finger vein identification method based on secondary screening and score fusion
Prasanth et al. Fusion of iris and periocular biometrics authentication using CNN
Jain et al. Multibiometrics
Arunkumar et al. PALMPRINT AND FACE BASED MULTIMODAL RECOGNITION USING PSO DEPENDENT FEATURE LEVEL FUSION.

Legal Events

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