KR100651734B1 - Method for multimodel biometric identification and system thereof - Google Patents

Abstract

The present invention relates to a multiple biometric method and a system thereof, the method of performing multiple biometrics using a plurality of single biometric systems, comprising: (a) a biometric of a user who has received an identification request from the single biometric system; Receiving a single biometric comparison value generated by comparing the information with biometric information about pre-registered users; (b) generating a comparison value vector for each registered user through the single biometric comparison value; (c) verifying whether the comparison value vector is a comparison value vector generated by the comparison of users for each registered user, and generating a determination value; And (d) generating a candidate list that lists, in a predetermined manner, a list of users most likely to be the user to be identified using the determination value. Therefore, more accurate user recognition is possible by using various biometric features.

Description

Method for multimodel biometric identification and system

1 is a block diagram of a multiple biometric system in accordance with the present invention.

2 is a flowchart of a multiple biometric method according to the present invention.

3 is a block diagram associated with training a binary classifier in FIG.

FIG. 4 is a diagram illustrating an example of training a binary classifier in FIG. 3.

FIG. 5 is a diagram illustrating an example of a discrimination boundary line found in FIG. 4.

FIG. 6 is a diagram illustrating another example of the discrimination boundary line found in FIG. 4.

7 is a flowchart illustrating a method of learning a binary classifier in FIG. 1.

The present invention relates to multiple biometric methods and systems thereof, and more particularly, to multiple biometric methods and systems for performing multiple biometrics using comparison values of a plurality of single biometric systems.

The biometric system is a system for identifying an identity using biometric information of a user. More specifically, the biometric system may be classified into verification and identification.

Authentication is a way of verifying identity through a one-to-one (1: 1) comparison, where a user is verified as identity, while recognition is one-to-many (1: many) with biometric data of several people registered in a database. A comparison is a way to find out who is most similar and identify you. That is, authentication is generated with a binary class (two class) result of yes / no form, whereas recognition is generated with a list of candidates in order of most likely candidates.

The features used in biometrics include physical features such as face, fingerprint, vein and iris, and behavioral features such as signature, gait and voice. Single biometrics is a method of using only one of these biometric features of a user. However, face recognition is sensitive to light changes, and fingerprints have many advantages and disadvantages, depending on the characteristics of the body used. . In particular, since there is a limit to the number of cases in which one biometric feature can be expressed, that is, the degree of freedom of expression of the biometric feature, a biometric system in which a large number of users are registered inevitably has a high performance and reliability with only one biometric feature. It becomes difficult to implement. Therefore, by performing user recognition using various biocharacteristics, the performance and reliability of the biometric system can be improved.

However, since the range and form of the value of the biometric data and the proper recognition method are different according to the biocharacteristic, there are many difficulties in implementing a biometric system by processing various biometric data at once. And, unlike recognition, since the candidate list in the form of a multi class is generated as a result, it is difficult to make a single candidate list by integrating multiple single biometric results including many candidates. .

In order to solve the problems of the prior art, the present invention provides a multi-biometric system using a plurality of single biometric results regardless of the type and number of biometric features used and the number of registered users generated in a single biometric system. It provides a method and system that can be efficiently implemented.

The multi-biometric method of the present invention for solving the above technical problem, in a method of performing a multi-biometric using a plurality of single biometric system, (a) the identification request from the single biometric system Receiving a single biometric comparison value generated by comparing biometric information of a user and biometric information of registered users; (b) generating a comparison value vector for each registered user through the single biometric comparison value; (c) verifying whether the comparison value vector is a comparison value vector generated by the comparison of users for each registered user, and generating a determination value; And (d) generating a candidate list which lists, in a predetermined manner, a list of users most likely to be the user to be identified using the determination value.

Multi-biometric system of the present invention for solving the above technical problem, in a system for performing multiple biometrics using a plurality of single biometric system, the biometric of the user who received the identification request from the single biometric system An input unit configured to receive a single biometric comparison value generated by comparing the information with biometric information about pre-registered users; A comparison value vector generation unit for generating a comparison value vector for each registered user through the single biometric comparison value; A binary classification unit for generating a discrimination value by verifying whether the comparison value vector is a comparison value vector generated by a comparison of users for each registered user; And a recognition list generator configured to generate a candidate list listing a list of users most likely to be the users who are to be identified using the determination value in a predetermined manner.

The present invention also provides a computer-readable recording medium having recorded thereon a program for executing the multiple biometric method of the present invention for solving the above technical problem.

Hereinafter, with reference to the accompanying drawings will be described in detail a multi-biometric method and system according to the present invention.

1 is a block diagram of a multiple biometric system in accordance with the present invention. Referring to FIG. 1, the multi-biometric system includes a single biometric system 100, a normalizer 120, a comparison vector generator 140, a binary classifier 160, and a recognition list generator 180. do.

The single biometric system 100 is comprised of a first single biometric system 101, a second single biometric system 102 and a third single biometric system 103. Biometric information of the user requesting identification is input to each of the first single biometric system 101, the second single biometric system 102, and the third single biometric system 103.

Each single biometric system (101) (102) (103) displays biometric information of a user requesting identification and biometric information of users already registered in each single biometric system (101) (102) (103). Compare.

으로 나타난다. 여기에서,

은 신원 확인을 요청하는 사용자의 생체정보와 제 1 단일생체인식 시스템(101)에 등록된 n번째 사용자의 생체정보를 상호 비교하여 생성된 생체정보 비교값이다.More specifically, the first single biometric system 101 compares biometric information of a user requesting identification with biometric information for each of the users (n persons) already registered in the first single biometric system 101. The biometric information comparison value is generated for each registered user. The biometric information comparison value generated by the first single biometric system 101 is

Appears. From here,

Is a biometric information comparison value generated by comparing biometric information of a user requesting identification with biometric information of an nth user registered in the first single biometric system 101.

으로 나타난다. 여기에서,

은 신원 확인을 요청하는 사용자의 생체정보와 제 2 단일생체인식 시스템(102)에 등록된 n번째 사용자 의 생체정보를 상호 비교하여 생성된 생체정보 비교값이다.Similarly, the second single biometric system 102 compares biometric information of a user requesting identification with biometric information for each of the users (n persons) already registered in the second single biometric system 102. A biometric information comparison value is generated for each registered user. The biometric information comparison value generated by the second single biometric system 102 is

Appears. From here,

Is a biometric information comparison value generated by comparing biometric information of a user requesting identification with biometric information of an nth user registered in the second single biometric system 102.

으로 나타난다. 여기에서,

은 신원 확인을 요청하는 사용자의 생체정보와 제 3 단일생체인식 시스템(103)에 등록된 n번째 사용자의 생체정보를 상호 비교하여 생성된 생체정보 비교값이다.Similarly, the third single biometric system 103 compares biometric information of a user requesting identification with biometric information for each of the users (n persons) already registered in the third single biometric system 103. A biometric information comparison value is generated for each registered user. The biometric information comparison value generated by the third single biometric system 103 is

Appears. From here,

Is a biometric information comparison value generated by comparing biometric information of a user requesting identification with biometric information of an nth user registered in the third single biometric system 103.

The normalization unit 120 normalizes the biometric information comparison values input from each of the single biometric systems 101, 102, and 103 to have a common range and a unit value. As such, the reason for performing normalization through the normalization unit 120 is that the comparison values of the biometric information of some single biometric systems 100 are shown in the form of similarity, and the remaining single biometric systems 100 are compared in the form of dissimilarity. It is to unify it into any one. In addition, when the comparison of the biometric information of each single biometric system (101, 102, 103) has a different range of values, it is to unify it into a common range of values.

More specifically, for example, a value in a common range refers to a value in a range such as 0 to 1 or 0 to 100, which is a range more easily recognized by a user. However, normalizing values in a common range such that 0 to 1 or 0 to 100 are shown as an example, and the range of various ranges are used to facilitate the learning of the binary classification unit 160 and to improve the classification performance. Value can be used.

행렬로 비교값 벡터를 생성하는데, 신원 확인을 요청한 사용자의 생체정보와 단일 생체인식 시스템(100)에 등록되어 있는 1번째 사용자의 생체정보를 비교하여 생성된 비교값 벡터는 1번째 행에서

으로 나타나고, 2번째 사용자의 생체정보를 비교하여 생성된 비교값 벡터는 2번째 행에서

으로 나타나며, n번째 사용자의 생체정보를 비교하여 생성된 비교값 벡터는

번째 행에서

으로 나타난다.The comparison value vector generation unit 140 generates a comparison value vector for each registered user from the normalized biometric comparison value through the normalization unit 120. Here, when the biometric information comparison values generated and input in the single biometric system 100 have a common range and unit value, they may be directly input to the comparison value vector generator 140 without going through the normalization unit 120. have. The comparison value vector generator 140

A comparison value vector is generated using a matrix, and the comparison value vector generated by comparing the biometric information of the user who has requested identification with the biometric information of the first user registered in the single biometric system 100 is shown in the first row.

The comparison value vector generated by comparing the biometric information of the second user is shown in the second row.

The comparison value generated by comparing the biometric information of the nth user is

In the first row

Appears.

행렬로 판별값을 생성하는데, 단일 생체인식 시스템(100)에 등록되어 있는 1번째 사용자의 판별값은 1번째 행에서

으로 나타나고, 2번째 사용자의 판별값은 2번째 행에서

으로 나타나며, n번째 사용자의 판별값은

번째 행에서

으로 나타난다.The binary classification unit 160 uses the pre-trained data on the comparison value vectors generated for each registered user, whether the comparison value vector is generated by comparison or the comparison value vector is generated by comparison of others. This will be verified in turn. The final similarity (or dissimilarity) of comparing the reliability or the likelihood of the comparison value vector generated by the person comparison or the comparison value vector generated by the comparison of others through the discrimination value generated as the verification result. Indicates. Here, the determination value generated as a result of the verification by the binary classification unit 160 does not represent a digital value that is divided into '0' and '1', but an analog value that is continuously displayed as '0' to '1'. It represents any one of the values. More specifically, in the binary classification unit 160

The discrimination value is generated by the matrix, and the discrimination value of the first user registered in the single biometric system 100 is determined in the first row.

And the discriminated value of the second user is

The determination value of the nth user is

In the first row

Appears.

The recognition list generating unit 180 generates the recognition list in the order of high probability or low probability that the user has confirmed the identity request by multiple biometrics using the determination value of the binary classification unit 160 for each registered user. Create

In FIG. 1, the number of single biometric systems 100 is limited to three to illustrate the present invention more easily. However, the present invention is not limited to the type and number of biometric information used. In addition, a plurality of single biometric systems may be used for the same single living body using different recognition methods.

2 is a flowchart of a multiple biometric method according to the present invention. Referring to FIG. 2, first, biometric information of a user to be identified is input to a plurality of single biometric systems 100 (S200).

Next, biometric information comparison values are received from each of the plurality of single biometric systems 100 (S210). Here, the biometric information comparison value refers to a value generated by comparing biometric information of a user requesting identification with biometric information of users already registered in each single biometric system 100.

Next, the comparison of the biometric information input in step S210 is normalized (S220). Normalization refers to the biometric comparison values input from each single biometric system 100 to have a common range and unit value. The reason for performing normalization as described above is that the biometrics comparison values of some single biometric systems 100 appear in the form of similarity, and the other single biometric systems 100 represent the comparison values in the form of dissimilarity. To do so. In addition, when the comparison of the biometric information of each single biometric system (101, 102, 103) has a different range of values, it is to unify it into a common range of values.

Next, a comparison value vector is generated for each user registered in each single biometric system 100 with respect to the normalized biometric information comparison value (S230). Here, when the biometric information comparison values generated and input from the single biometric system 100 from step S210 have a common range and unit value, the process may directly proceed to step S230 without passing through step S220.

Next, with respect to the comparison value vectors generated for each registered user in step S230, it is determined whether the comparison value vector generated by the person comparison or the comparison value vector generated by the comparison of others using the pre-trained binary classification unit. In order to verify it in turn generates a determination value (S240). Here, the discrimination value represents the final similarity (or dissimilarity) in which the reliability or possibility of being a comparison value vector generated by the person comparison or the comparison value vector generated by the comparison of others is compared. The discrimination value does not represent a digital value that is divided into '0' and '1', but represents a value of any one of analog values continuously displayed as '0' to '1'.

Next, using the size of the determination value generated in step S240 to generate a list of recognition in the order of registered users most likely to be the user who wants to identify the identity (S250).

3 is a block diagram associated with training a binary classifier in FIG. Referring to FIG. 3, when there are n users registered in the single biometric system 100, the comparison value vector generation unit 140 from the binary classification unit 160 to one comparison value vector based on comparison with others and others N-1 comparison value vectors by comparison are input. In FIG. 3, the binary classification unit 160 is trained through the biometric information on the second user registered in the single biometric system 100. That is, one comparison value vector based on self comparison and one n-1 comparison value vector based on another person's comparison are input to the binary classification unit 160, and the binary classification unit 160 analyzes the corresponding comparison vector based on this. To do learning.

The learning result binary classifier 160 may classify whether the comparison value vector input from the comparison value vector generation unit 140 is a comparison value vector generated by comparison with others or a comparison value vector generated by comparison with others. Will be. Parts not described in FIG. 3 will be referred to FIG. 1.

FIG. 4 is a diagram illustrating an example of training a binary classifier in FIG. 3. Referring to FIG. 4, a comparison value vector generated by using a face and a fingerprint biometric information is displayed on a two-dimensional plane.

The blue dot indicates a comparison value vector generated by Genuine, and the red x mark indicates a comparison value vector generated by Impostor. Here, learning a binary classifier means finding a discriminating boundary so that a vector generated by a comparison of oneself and a vector generated by a comparison of others can be distinguished well.

FIG. 5 is a diagram illustrating an example of a discrimination boundary line found in FIG. 4. Referring to FIG. 5, an example of a linear discrimination boundary line found using data given for learning is shown.

Here, the position and type of the linear discrimination boundary line are determined according to the data used and the learning algorithm of the binary classification unit, and the present invention is not limited to a specific binary classification algorithm. In addition, the data given for learning means data whose membership is known whether each vector is generated by comparison with others or by comparison with others.

Therefore, the discrimination value is calculated using the distance between the linear discrimination boundary line and the comparison value vector. In other words, a comparison vector farther to the blue from the discrimination boundary is more likely to be generated by the comparison of the person, and on the contrary, a comparison vector farther to the red is more likely to be generated by the comparison of others.

Therefore, when actually used in a multi-biometric system, when a random comparison vector that does not know whether it belongs to the comparison value vector generated by the comparison of yourself or the comparison vector generated by the comparison of others, the center of the discrimination boundary This can be determined.

Here, the present invention has been described with a single biometric information of a face and a fingerprint for convenience of the detailed description of the present invention. However, the present invention is not limited thereto and may be extended to using three or more single biometric information.

FIG. 6 is a diagram illustrating another example of the discrimination boundary line found in FIG. 4. Referring to FIG. 6, an example of a nonlinear discrimination boundary line found using data given for learning is shown.

In FIG. 5, the linear discrimination boundary line is illustrated, but in FIG. 6, except for the nonlinear discrimination boundary line, the remaining undescribed parts will be described with reference to FIG. 5.

7 is a flowchart illustrating a method of learning a binary classifier in FIG. 1. Referring to FIG. 7, first, the binary classification unit 160 receives a comparison value vector generated by comparison with another and a comparison value vector generated by tine comparison from the comparison value vector generator 140 (S700).

Next, a training database and a test database are constructed through the comparison value vector generated by the comparison of the person received in step S700 and the comparison value vector generated by the comparison of others (S710).

Next, the binary classification unit 160 is trained using the learning database built in step S710 (S720). Through such learning, the binary classification unit 160 may classify the comparison value vector generated by comparing the person and the comparison value vector generated by comparing the other person. The discriminant value in the trained binary classifier 160 indicates the reliability of the possibility of the vector generated by the comparison of the users.

Next, the learning result is verified through the result of inputting the comparison value vector stored in the test database for the binary classification unit 160 trained in step S720 to the binary classification unit 160 (S730). .

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD_ROM, magnetic tape, floppy disks, and optical data storage, and may also include those implemented in the form of carrier waves (eg, transmission over the Internet). . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, optimal embodiments have been disclosed in the drawings and the specification. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the invention as defined in the claims or the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, by presenting a method of implementing multiple biometric systems using a single biometric result, multiple biometric methods and systems having higher performance and reliability can be configured.

In addition, the multi-biometric method and system using a single biometric result proposed in the present invention simplify a single biometric result in a multi-class form into a binary class problem, thereby easily integrating a large number of single biometric systems. You can do it.

In addition, it is possible to construct multiple biometric methods and systems that can represent multiple single biometric results in one integrated comparison, regardless of the number of candidates generated in a single biometric system.

Claims (17)

In the method of performing multiple biometrics using a plurality of single biometric systems, (a) a plurality of single biometric comparison values generated by comparing the plurality of biometric information of a user who has received an identification request with multiple biometric information of previously registered users in each of the plurality of single biometric systems. Receiving input from each of the recognition systems; (b) collecting a single biometric comparison value input from each of the plurality of single biometric systems for each registered user and generating a comparison value vector for each registered user; (c) verifying whether the comparison value vector is a comparison value vector generated by comparison with the user, and generating a determination value for each registered user; And and (d) generating a candidate list which lists, in a predetermined manner, a list of users most likely to be the user whose identity is to be identified using the determination value. The method of claim 1, wherein step (a) (a-1) transmitting, to each of the plurality of single biometric systems, multiple biometric information about a user who has received the identification request; And (a-2) receiving a single biometric comparison value from each of the plurality of single biometric systems. The method of claim 2, The single biometric comparison value is a value that represents a numerical value indicating a degree of mutual matching by comparing biometric information about a user who has received the identification request with biometric information about users who are pre-registered in the plurality of single biometric systems. Multiple biometric methods characterized by. The method of claim 1, wherein step (b) (b-1) normalizing the single biometric comparison value; And and (b-2) setting the normalized single biometric comparison value into a comparison value vector for each registered user. The method of claim 4, wherein Normalizing the single biometric comparison value in the step (b-1), If the single biometric comparison value does not match the similarity and dissimilarity, so as to match either one of the similarity or dissimilarity. The method of claim 4, wherein Normalizing the single biometric comparison value in the step (b-1), Multiple biometric methods such that the range is matched if the single biometric comparison value does not match the range. The method of claim 1, wherein step (c) (c-1) verifying whether the comparison value vector is a comparison value vector based on comparison of oneself or a comparison value vector based on comparison of others using a binary classification unit; And (c-2) generating a discrimination value between the comparison value vector based on the comparison of the person and the comparison value vector based on the comparison of other persons through the verification in the step (c-1); . The method of claim 7, wherein The binary classification unit is a multi-biometric method, characterized in that previously learned. The method of claim 8, wherein the binary classifier is pre-learned. Generating a learning database by collecting an input vector generated by comparison of persons and an input vector generated by comparison of others using a single biometric result generated by the single biometric system; Generating a test database by collecting an input value vector generated by comparing a person and an input value vector generated by comparing a person using a single biometric result generated by the single biometric system; Inputting a comparison value vector based on comparison of oneself and a comparison value vector based on comparison of another person in the learning database by inputting the binary classification unit; And Verifying a learning result of the binary classification unit using the test database. 2. In a system for performing multiple biometrics using a plurality of single biometric systems, Each of the plurality of single biometric systems includes a single biometric comparison value generated by comparing the multiple biometric information of the user who has received the identification request with the multiple biometric information of the registered users in each of the plurality of single biometric systems. An input unit to receive from the input unit; A comparison value vector generation unit for collecting a single biometric comparison value input from each of the plurality of single biometric systems for each registered user and generating a comparison value vector for each registered user; A binary classification unit for verifying whether the comparison value vector is a comparison value vector generated by comparison of users and generating a determination value for each registered user; And And a recognition list generator configured to generate a candidate list for arranging a list of users most likely to be the users to be identified using the determination value in a predetermined manner. The method of claim 10, The single biometric comparison value is a value that represents a numerical value indicating a degree of mutual matching by comparing biometric information about a user who has received the identification request with biometric information about users who are pre-registered in the plurality of single biometric systems. Multiple biometric systems characterized. The method of claim 10, Further comprising a normalization unit for normalizing the single biometric comparison value, The comparison value vector generating unit generates a comparison value vector for each registered user through the normalized biometric comparison value. The method of claim 12, wherein normalizing the single biometric comparison value comprises: And if the single biometric comparison value does not match the similarity or dissimilarity, it matches the one of the similarity or dissimilarity. The method of claim 12, wherein normalizing the single biometric comparison value comprises: And wherein the single biometric comparison values match the ranges if the ranges do not match. The method of claim 10, wherein the binary classification unit, The comparison value vector is verified by using a binary classification unit to verify whether the comparison value vector based on the comparison of the person or the comparison value vector based on the comparison of others, and generates a discrimination value. And generating a discrimination value according to the comparison value vector. The method of claim 15, The binary classification unit is a multi-biometric system, characterized in that pre-learned. A computer-readable recording medium having recorded thereon a program for performing the steps according to any one of claims 1 to 9.

Images