JP2008299589A - Biometrics system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the convenience of composite authentication by reducing the number of times of input of biological information by a user. <P>SOLUTION: In this system for comparing a plurality of biological information input on the basis of a prescribed input sequence with a prescribed registration template to calculate collation scores, and for integrally authenticating a principal by using successive probability ratio verification to the collation scores of the biological information input until that time, the distribution (principal distribution) of the collation scores of the biological information of the principal and the distribution (others' distribution) of the collation scores of the biological information of the principal and the others are searched for every user for every type of biological information by using a distribution estimation function 115, and a Kullback-Leibler distance is calculated from the principal distribution and the others' distribution for every type of biological information by using an inter-distribution distance calculation function 116, and the input sequence of the plurality of types of biological information is determined in the order of the larger Kullback-Leibler distance by using an input sequence optimization function 117. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a biometric authentication system, and more particularly to an authentication method, system, and program for authenticating a person using a plurality of biometric information of users.

  Biometric authentication has been put to practical use as an authentication method that has the advantage that it is difficult to forge compared to authentication based on a password, an IC card, and the like, and is not forgotten like a password. However, as a result of collating the biometric information acquired at the time of authentication with a template registered in advance, there may be a case where the person is not correctly authenticated or the other person is mistakenly authenticated with the person. The probability that the former occurs is called FRR (False Rejective Rate), and the probability that the latter occurs is called FAR (False Acceptance Rate).

In order to improve authentication accuracy, composite authentication has been proposed in which a plurality of pieces of biometric information (for example, fingerprints, irises, voiceprints, etc.) are input and the identity determination is integrated using these.
For example, in Patent Document 1 (Japanese Patent Application Laid-Open No. 2003-186836), biometric information to be input is arranged in order from the person who is likely to be authenticated based on the past use history of the user, and biometric information is used for authentication. There is disclosed a composite authentication in which collation and determination are sequentially performed each time the password is input, and the authentication is terminated when sufficient confidence is obtained. Thereby, in composite authentication, the number of input of biometric information necessary for authentication can be reduced, and convenience is improved.

More specifically, based on the past usage history of the user first, as FAR for each biometric information becomes full biometric information common target performance value FA _L, the threshold matching score of each biometric information Set the value. Next, the input order of the biological information is determined so that the FRR at the threshold value is in ascending order. In personal authentication, verification and determination are repeated each time one piece of biometric information is input, and the authentication ends when sufficient confidence is obtained. Determination, obtains a biometric information input FAR corresponding to the matching score (the number of inputs and i when) (FA (i)), and compares the with FA _L which (FA (i) ≦ FA If it is _L, it authenticates with the person and ends; OR determination) or FA (i) is integrated, and the integrated value is compared with a target performance value that is a function of the number of inputs i. The latter compares, for example, FA (1) × FA (2) ×... × FA (i) and (FA _L ) ⁱ (FA (1) × FA (2) ×... × FA (i) ≦ _(FA ^{L) i} if person and ending with authentication; FAR product determination) performed by.

  Non-Patent Document 1 discloses a method of performing composite authentication using a sequential probability ratio test. When the order of inputting a plurality of pieces of biometric information is determined, the average number of inputs of biometric information can be minimized by applying the sequential probability ratio test to the composite authentication.

JP 2003-186836 A Kenta Takahashi et al., “Multimodal Biometric Authentication Using Sequential Probability Ratio Test and Logistic Regression”, Science D Vol.J89-D No.5 pp.1061-1065 (2006) Takao Murakami et al., “Automatic Recognition of Japanese Vowel Sequences Using Structural Representation of Speech”, IEICE Technical Report SP2005-14 pp.13-18 (2005)

However, the method disclosed in Patent Document 1 has room for improvement in terms of minimizing the average number of times biometric information is input. For example, when performing personal authentication by OR determination, the verification score of the previous authentication stage is not used at all for the personal authentication in each authentication stage. Thus, for example, in i-1 th authentication stage, FA (i-1) is also as had slightly above FA _L, it is not it is considering in the authentication phase of the i-th and later, FA (i-1) is treated the same as in the case of greatly exceeded the FA _L. Even when the determination is made using the integrated value of the FAR value such as the FAR product, there is no theoretical guarantee that the average number of times of input of the biological information is minimized.

  In addition, when the order of input of a plurality of pieces of biological information disclosed in Non-Patent Document 1 is determined, the sequential probability ratio test is applied to composite authentication, so that biometric information can be compared with the method of Patent Document 1. The average number of inputs can be reduced. However, it is assumed that the input order of biometric information (for example, face → vein → fingerprint) is freely selected by the user, and there is a problem that the number of inputs increases depending on the order. For this reason, it is a problem to further improve convenience by optimizing the number of times of input including the order of input.

In addition, in the methods according to Patent Document 1 and Non-Patent Document 1, when the user's past usage history is small, there is a risk that the estimation accuracy of the identity distribution used for calculating the FRR and the distribution of others used for calculating the FAR are low. In particular, for biometric information in which the input order is positioned backward, there is a high possibility that the user's past usage history is remarkably reduced. As a result, there is a possibility that the biometric information may not be optimized so that the order of biometric information is easily authenticated, which hinders the reduction in the number of times biometric information is input.
These are problems in minimizing the average number of times the user's biometric information is input, and conventional authentication using a plurality of biometric information has room for improvement in terms of convenience.

  An object of the present invention is to improve the convenience of composite authentication by reducing the number of inputs of biometric information.

The biometric authentication system according to the present invention preferably includes one or a plurality of sensors that acquire a plurality of types of biometric information of the user, and a processing device that performs a user authentication process using the plurality of biometric information obtained from the sensors. A biometric authentication system comprising:
A database for registering the biometric information of the user obtained from the sensor as a template;
The processing apparatus obtains the biometric information of the user obtained from the sensor by comparing the biometric information of the user with the template of the user registered in the database, and the biometric information collating unit. From the collation score for the plurality of types of biometric information obtained, identity determination means for performing identity determination in an integrated manner using a sequential probability ratio test, and for each of the plurality of types of biometric information, for each user, A distribution estimation unit for obtaining a distribution of biometric information collation scores (personal distribution) and a biometric information collation score between the person and another person (other person distribution), and a plurality of types of the biometric information obtained by the distribution estimation unit. An inter-distribution distance calculation means for obtaining a Cullback libra distance from the principal distribution and the other person distribution; and The input order optimization means for determining the input order of the plurality of types of biological information so that the order of the Ibra distance is large, and the input order is finalized based on the input order determined by the input order optimization means Configured as a biometric authentication system comprising: an input order determining means for determining the biometric information; and an input biometric information indicating means for instructing the biometric information to be input in accordance with the input order finally determined by the input order determining means Is done.

In a preferred example, the distribution estimation means estimates the identity distribution and the others distribution using the past usage history of the user.
Preferably, the distribution estimation means estimates the individual distribution and the other person distribution by using the past use history other than the user himself / herself in addition to the past use history of the user himself / herself.
Preferably, the input order determination means changes the input order of the plurality of pieces of biological information of the user initially determined by the input order optimization means, and the input biological information instruction means changes the input Present the order to the user.
In one example, the input order determining means finally determines the input order of the plurality of biological information based on the random number generated by the pseudo random number generator.

The biometric authentication processing method according to the present invention is preferably a biometric authentication processing method for performing authentication processing using a plurality of types of biometric information of a user obtained from a sensor, wherein the user's biometric authentication processing method is obtained from the sensor. The biometric information is compared with the user's template registered in advance in the database to calculate a collation score, and the collation score for a plurality of types of the biometric information obtained by the biometric information collation is sequentially determined. A person determination step for performing a person determination in an integrated manner using a probability ratio test, and a plurality of types of biometric information for each user; A distribution estimation step for obtaining a distribution of other people's biometric information collation scores (others distribution), and a book for a plurality of types of the biometric information obtained by the distribution estimation An inter-distribution distance calculating step for obtaining a cullback / libra distance from the distribution and the other person's distribution, and inputting a plurality of types of the biological information so that the cullback / libra distance calculated by the inter-distribution distance calculation is in descending order. The biometric authentication method includes an input order determining step for determining the order and an input biometric information instruction step for instructing the biometric information to be input based on the determined input order.
Moreover, the biometric authentication program according to the present invention is configured as a program for executing the above means in the authentication system by a CPU.

According to the present invention, since the average number of biometric information inputs can be reduced from both the identity determination algorithm and the biometric information input order, the convenience of biometric authentication is improved.
In addition, by changing the input order of a plurality of biometric information determined at the beginning, the update frequency of the user's past usage history for the biometric information whose input order is positioned behind increases, The estimation accuracy is improved.

  Furthermore, the estimation accuracy of the personal distribution and the other person distribution is further improved by performing the distribution estimation using the past use history other than the user himself / herself. As a result, the input order of biometric information can be determined more appropriately, the number of inputs by the user can be further reduced, and convenience is further improved.

Hereinafter, embodiments will be described with reference to the drawings.
FIG. 1 shows a configuration example of a biometric authentication system.
This system is a composite authentication system having a function of optimizing the input order of biometric information, and a plurality of input sensors 1001 to 100N (these can be obtained) that can acquire different biometric information (biometric information 1 to N), respectively. And a terminal 110 that performs composite authentication and biometric information input order optimization processing using the acquired biometric information.

  The plurality of pieces of biometric information may have different modalities such as fingerprints, irises, and voiceprints, or may be configured with different parts of the same modality such as index finger fingerprints, middle finger fingerprints, and ring finger fingerprints. Alternatively, a mixture of both may be used. Here, modality refers to the type of biological information that can be acquired by one sensor. When a plurality of pieces of biological information are configured with different parts having the same modality, one input sensor is sufficient. The input sensor 100 is connected to the terminal 110 by a USB cable or the like.

  The terminal 110 includes an input order determination function 111 for finally determining the order of input biometric information to be presented by the user, an input biometric information instruction function 112 for instructing the user of biometric information to be input according to the order, and an input sensor. The biometric information collation function 113 that calculates the biometric information collation score acquired from 100, and the personal identification function 114 that calculates the cumulative score of the biometric information acquired so far using the collation score obtained there, and sequentially performs the individual determination. A distribution estimation function 115 for estimating the person distribution and the other person distribution for optimizing the input order, and an inter-distribution distance calculation function for calculating a distance between the person distribution estimated by the distribution estimation function 115 and the other person distribution. 116, an input order optimization function 117 for optimizing the input order using the inter-distribution distance calculated by the inter-distribution distance calculation function 116, and a database (DB) consists of 118.

The DB 118 holds user registration data 120 for each user. The user registration data 120 includes a user ID 121, template information 122 in which one or a plurality of registration templates are stored for each piece of biometric information, a matching score with the person and a matching score with another person for each piece of biometric information Score information 123 stored therein, and (o ₁ , o ₂ ,..., O _N ) input order information 124 storing biometric information numbers in each order. Here, o _i is the number of biometric information to be input for the i-th time (1 ≦ o _i ≦ N). For example, when biometric information 4 is input first time, biometric information 3 is input _{second time} , and so on, o ₁ = 4, o ₂ = 3, and so on. FIG. 6 schematically shows user registration data for each user.

FIG. 2 shows a hardware configuration of the terminal 110 in the biometric authentication system. The terminal 110 includes a CPU 200, a memory 201, an HDD 202, an input device 203, and an output device 204. Here, the input device 203 is a plurality of input sensors 100, and the output device 204 has a display for instructing the user to input.
The DB 118 is formed in the HDD 202. In addition, the memory 201 stores an application program for biometric authentication processing according to the present embodiment, and when the program is executed by the CPU 200, the above functions 111 to 117 are realized.
In this example, the terminal 110 is referred to, but this may be a computer having the above configuration or a small device.

FIG. 3 shows a biometric authentication process and data flow.
The terminal 110 initializes the input count i of the biometric information to 1 and the initial value Z ₀ of the cumulative score to 1 (S301).
Next, the terminal 110 receives an input of the user ID 121 (S302). Note that the user ID 121 may be input using an IC card or a keyboard operation.

The terminal 110 acquires the input order information 124 of the input user ID 121 (hereinafter referred to as input user ID) from the DB 118. Based on the input order information 124, the order of the input biometric information to be presented by the user is determined. At this time, the order is sometimes determined to be different from the input order information 124 (S303). As a method of changing the initially determined input order information 124, for example, using a pseudo random number generator, a random order (p ₁ , p ₂ ,..., Unrelated to the input order information 124 with a certain probability). There is a method of generating p _N ) (1 ≦ p _i ≦ N, 1 ≦ i ≦ N) and determining it according to the order.

In the prior art, much less that the score information 123 of the biological information are located is updated to the rear in the input sequence of the plurality of biometric information, for the data shortage, person distribution estimates at S307 f _G ( There is a problem that the estimation accuracy of s _i ) and the distribution of others f _I (s _i ) is low. As described above, sometimes the order different from the input order information 124 is used to increase the update frequency of the score information 123 of the biological information positioned behind, and the personal distribution f _G (s _i ) and the other person distribution f _I. The estimation accuracy of (s _i ) is improved. As a result, the input order of biometric information can be determined more appropriately (a method for determining the input order will be described later). Thereby, the user can further reduce the number of inputs, and convenience is further improved.

According to the determined input sequence instructs the user by displaying on the display to enter the biometric information o _i (S304). Then, the terminal 110 receives an input of biometric information _{o i} (S305).
Terminal 110, the input user ID of the template information 122 in the DB 118, obtains the registration template of the biometric information o _i, the input biometric information (hereinafter, referred to as input biometric information) matches the registered template and, matching score s _i is calculated (S306). When there are a plurality of registered templates, any one of them may be selected as the matching score s _i . Or all by calculating the average of the match score for may be it as a matching score s _i.

Terminal 110 of the score information 123 of the input user ID acquired from the DB 118, log likelihood ratio using the matching score corresponding to the biometric information _{o i} person distribution _f G (s), others distribution _f I (s) log (f _G (s) / f _I (s)) is obtained (S307). A method for obtaining log (f _G (s) / f _I (s)) will be described in detail later.

The terminal 110 uses the matching score s _i and the log likelihood ratio log (f _G (s) / f _I (s)) to accumulate the score Z _i (Z _i ← Z _i−1 + log (f _G (s _i )). / F _I (s _i ))) is calculated (S308). The calculation method of Z _i is based on the sequential probability ratio test. The sequential probability ratio test will be described in detail later.

The terminal 110 determines whether or not the condition: (i <N Z _i ≧ _TG ) ∨ (i = N∧Z _i ≧ λ) is satisfied. If the determination value is true, the process proceeds to S313 as authentication OK. If the determination value is false, the process proceeds to S310 (S309). T _G and λ are threshold values used in the sequential probability ratio test. T _G and λ will be described in detail later.

The terminal 110 determines whether or not the condition: (i <N Z _i ≦ T _I ) ∨ (i = N∧Z _i <λ) is satisfied. If the determination value is true, authentication NG is assumed. If the determination value is false, the process proceeds to S311 (S310). T _I is a threshold value used in the sequential probability ratio test. It will be described in detail later T _I.
Next, the terminal 110 increments the biometric information input count i by 1 (S311), and returns to the process of S304 (S312).

Further, the terminal 110 updates the matching score with the person in the score information of the input user ID and the matching score with another person (S313). The collation score with the person is updated by adding all the collation scores obtained in S306 to the DB 118. If there is a registered template for which a matching score is not obtained in S306 (for example, when a matching score is obtained by selecting one from a plurality of registered templates), the matching score may be obtained here and added to the DB 118. . The update of the collation score with another person is performed by acquiring the template information 122 other than the input user ID from the DB 118, obtaining the collation score with each input biometric information input so far, and adding it to the DB 118.
The terminal 110 performs an input order optimization process on the input user ID (S314). A specific method of the input order optimization process will be described next.

FIG. 4 shows an input sequence optimization process and data flow.
The terminal 110 acquires score information 123 for all user IDs 121 from the DB 118. Using these, for each biometric information, the identity distribution f _G (s) and the stranger distribution f _I (s) of the user ID 121 (hereinafter referred to as optimization target user ID) that performs input order optimization are estimated (S401). The method for estimating the distribution will be described in detail later.

Next, the terminal 110 calculates the inter-distribution distance between the individual distribution f _G (s) and the other person distribution f _I (s) for each biological information (S402). A method for calculating the inter-distribution distance will be described in detail later.
Furthermore, the terminal 110 optimizes the input order of the biological information based on the inter-distribution distance, and sends the input order information to the database (S403). A method for optimizing the input order will be described in detail later.

  Next, the sequential probability ratio test in S306 will be described in detail. The sequential probability ratio test is based on the assumption that the observation data (matching score) is iid (independently follows the same distribution) and that a sufficient number of observations can be repeated. This is the optimal identity determination algorithm in the sense of minimizing.

FIG. 5 shows an authentication OK / NG area of the personal authentication algorithm based on the sequential probability ratio test when N = 3. In the sequential probability ratio test, the log likelihood ratio log (f _G (s _i ) / f _I (s _i )) of the person / others after observing the i-th matching score s _i is accumulated. The score Z _i is obtained (Z _i ← Z _{i -1} + log (f _G (s _i ) / f _I (s _i ))). This corresponds to obtaining the cumulative score Z _i as the log likelihood ratio of the person / others after observing the matching scores s ₁ ,..., S _i . When i <the N, it is determined that the person when Z _i exceeds the threshold T _G exits, and exit determination with others below the threshold T _I. When i = N, Z _i is compared with the threshold λ, and when Z _i is larger, the person is determined, and when Z _i is smaller, the person is determined as another person, and the process ends.

That is, if the condition: (i <N∧Z _i ≧ T _G ) ∨ (i = N∧Z _i ≧ λ) is satisfied, it is determined that the person is the person, and the process ends, and the condition: (i <N∧Z _i ≦ T _I ) If ∨ (i = N∧Z _i <λ) is satisfied, it is determined as another person and the process ends. In this way, by using the sequential probability ratio test, it is possible to perform composite authentication based on a person determination algorithm that minimizes the average number of times biometric information is input. As a result, the user can reduce the number of times of input, and an effect of improving convenience can be obtained.
log (f _G (s) / f _I (s)) may be obtained after estimating the distribution of f _G (s) and f _I (s) in the same framework as the distribution estimation method described below. Alternatively, the logistic regression described in Non-Patent Document 1 may be applied.

Next, the distribution estimation method in S401 will be described in detail.
First, a method for obtaining f _G (s) and f _I (s) for each biological information by maximum likelihood estimation will be described. A description will be given by taking f _G (s) as an example. Assuming that f _G (s) follows a normal distribution, it is only necessary to be able to estimate the average value, variance value μ _ML , and σ ² _ML . If the score information 123 of the optimization target user ID acquired from the DB 118 stores the matching scores s _G1 , s _G2 ,..., S _GK (assuming there are K pieces) with the person, ,
μ _ML = Σs _Gk / K
σ ² _ML = Σ (s _Gk −μ _ML ) ² / K
(1 ≦ k ≦ K).

A similar estimation can be made for f _I (s). Maximum likelihood estimation is a simple distribution estimation method and has advantages such as easy implementation and high processing speed. However, when the amount of data for distribution estimation (ie, K) is small, the estimation accuracy of the distribution is low. There is a problem.

Next, a method for obtaining f _G (s) and f _I (s) for each biological information by MAP (Maximum a Posteriori) estimation will be described. MAP estimation is known as one of the methods that makes it possible to estimate distribution more robustly than maximum likelihood estimation, even when data other than the user is used well, even when the amount of data of the user is small. . A description will be given by taking f _G (s) as an example. Assuming that f _G (s) follows a normal distribution, the MAP estimation calculates the average value and the MAP estimated values μ _MAP and σ ² _MAP of the variance values. From the optimization target user ID score information 123 acquired from the DB 118, the average value μ _ML and the variance value σ ² _ML are obtained by maximum likelihood estimation. For other user IDs 121 (assuming that there are M people), an average value μ _m and a variance value σ ² _m (1 ≦ m ≦ M) are obtained by maximum likelihood estimation for each user ID. At this time, if the method described in Non-Patent Document 2 is used,
μ _MAP = μ ′
σ ² _MAP = β ′ / α ′
Can be obtained as However,
μ ′ = (τμ + nμ _ML ) / (τ + n)
_{^{β'= β + nσ 2 ML /}} 2 + τn (μ ML -μ) 2/2 (τ + n)
α ′ = α + n / 2
τ = Σσ ² _m / Σ (μ _m −μ) ²
μ = Σμ _m / M
β = 1
α = M / Σσ ² _m
It is. The same can be obtained for f _I (s).

  By using MAP estimation as described above, the distribution can be estimated more robustly even when the amount of data obtained is small. As a result, the input order of biological information can be determined more appropriately. Thereby, the number of user inputs can be further reduced, and convenience is further improved.

Finally, the details of the method for calculating the distance between distributions and the method for optimizing the input order in S402 and S403 will be described.
First, as a distribution distance of the principal distribution _f G _{(s i)} with others distribution _f I _{(s i)} for each biometric information, Kullback-library distance _{∫f G (s i) log (} f G (s i) / f _I (s _i )) ds _i is determined. When f _G (s _i ) is a normal distribution with an average value μ _G and a variance value σ ² _G , and f _I (s _i ) is a normal distribution with an average value μ _I and a variance value σ ² _I , the Cullback library the distance is determined as _{-1 / 2-log (σ G} / σ I) + {σ 2 G + (μ I -μ G) 2} / 2σ 2 I.

Next, the input order of the biological information is arranged in ascending order of the value. The Cullback-libra distance is an expected value of the log likelihood ratio log (f _G (s _i ) / f _I (s _i )) of the person / others when the person inputs biometric information in the sequential probability ratio test. Since it has meaning, the value log (f _G (s _i ) / f _I (s _i )) that is sequentially added to the cumulative score Z _i by arranging the input order so that this value is in ascending order is as follows: They will be arranged in order from the largest to the smallest. Therefore, the average number of biometric information inputs is minimized. As a result, the user can reduce the number of times of input, and an effect of improving convenience can be obtained.

As described above, according to the composite authentication system of the present embodiment, the sequential probability ratio test is used as the identity determination algorithm, and further, the Cullback library distance between the identity distribution f _G (s _i ) and the other person distribution f _I (s _i ) is It is possible to optimize the input order of the biological information so that the order is larger from the smallest. Incidentally, in the conventional composite authentication system, it was not based on a framework that theoretically minimizes the average number of inputs of biometric information. However, according to the present embodiment, both the identity determination algorithm and the input order of biometric information Therefore, the average number of input of biometric information can be minimized. As a result, the user can reduce the number of inputs, and convenience is improved.

Moreover, according to the composite authentication system of the present embodiment, it is possible to change the optimized input order from time to time and instruct the user to input biometric information according to the changed input order. Incidentally, in the prior art, the biometric information score information 123 positioned backward in the input order is very rarely updated, and the amount of data is insufficient, so that the person distribution f _G (s _i ) and the other person distribution f _I There was a problem that the estimation accuracy of (s _i ) was low. According to the present embodiment, the update frequency of the score information of the biological information positioned behind increases, and the estimation accuracy of the individual distribution f _G (s _i ) and the other person distribution f _I (s _i ) is improved. Furthermore, according to the present embodiment, the distribution estimation can also be performed using the past use history other than the user himself / herself, and the estimation accuracy of the distribution is further improved. As a result, the input order of biometric information can be determined more appropriately, and the user can further reduce the number of inputs, further improving convenience.

The authentication according to the present embodiment can be applied to any application that performs user authentication based on biometric information. For example, it can be applied to information access control in an in-house network, identity verification in an Internet banking system or ATM, login to a member website, personal authentication when entering a protected area, login of a personal computer, and the like.

The figure which shows the structural example of the biometrics authentication system in one Example. The figure which shows the hardware constitutions of the biometrics authentication system by one Example. The flowchart which shows the biometrics authentication process in one Example. The flowchart which shows the input order optimization process in one Example. The figure which shows the outline of the authentication algorithm based on the sequential probability ratio test in one Example. The figure which shows the example of the user registration data of DB118 in a biometrics authentication system.

Explanation of symbols

100: Biological information input sensor 110: Terminal
111: Input order determination function 112: Input biometric information presentation function
113: Biometric information collation function 114: Identity determination function
115: Distribution estimation function 116: Inter-distribution distance calculation function
117: Input order optimization function 118: Database 120: User registration data 121: User ID
122: Template information 123: Score information 124: Input order information 200: CPU
201: Memory 202: HDD
203: Input device 204: Output device

Claims (9)

A biometric authentication system comprising one or more sensors for acquiring a plurality of types of biometric information of a user and a processing device that performs a user authentication process using a plurality of biometric information obtained from the sensors,
A database for registering the biometric information of the user obtained from the sensor as a template;
The processor is
Biometric information collating means for calculating a collation score by comparing the biometric information of the user obtained from the sensor with the template of the user registered in the database;
A person determination means for performing a person determination in an integrated manner using a sequential probability ratio test from the comparison scores for a plurality of types of the biological information obtained by the biological information collation means;
Distribution estimation means for obtaining, for each of a plurality of types of biometric information, a collation score distribution between the biometric information of the user (personal distribution) and a distribution of the collation scores of the biometric information of the user and the other person (distribution of other people). When,
An inter-distribution distance calculating means for obtaining a cullback / libra distance from the principal distribution and the other person distribution with respect to a plurality of types of biological information obtained by the distribution estimating means;
An input order optimizing unit that determines an input order of the plurality of types of the biological information so that the distance between the distributions is calculated in the descending order.
An input order determination means for finally determining an input order based on the input order determined by the input order optimization means;
A biometric authentication system comprising: input biometric information instruction means for instructing the biometric information to be input according to the input order finally determined by the input order determination means.   The biometric authentication system according to claim 1, wherein the distribution estimation unit estimates the identity distribution and the others distribution using a past usage history of the user.   The distribution estimation unit estimates the personal distribution and the other person distribution by using a past usage history other than the user himself / herself in addition to the past usage history of the user himself / herself. Biometric authentication system.   The input order determining means changes the input order of the plurality of pieces of biometric information of the user initially determined by the input order optimization means, and the input biometric information instruction means presents the changed input order to the user The biometric authentication system according to any one of claims 1 to 3, wherein: The biometric authentication system according to claim 4, wherein the input order determining unit changes so as to increase the update frequency of the score information of the biometric information positioned behind the plurality of biometric information of the user. 5. The biometric authentication system according to claim 4, wherein the input order determining means determines an input order of the plurality of pieces of biometric information based on a random number generated by a pseudo random number generator. A biometric authentication processing method for performing an authentication process using a plurality of types of biometric information of a user obtained from a sensor,
A biometric information matching step of calculating a matching score by comparing the biometric information of the user obtained from the sensor with the template of the user registered in advance in a database;
An identity determination step for performing identity determination in an integrated manner using a sequential probability ratio test from the verification scores for a plurality of types of the biological information determined by the biometric information verification;
A distribution estimation step for obtaining, for each of the plurality of types of biometric information, a collation score distribution between the biometric information of the user (personal distribution) and a collation score distribution of the biometric information of the principal and the other person (distribution of other people). When,
An inter-distribution distance calculating step for obtaining a Cullback libra distance from the principal distribution and the other person distribution for the plurality of types of biological information obtained by the distribution estimation;
An input order optimizing step for determining an input order of a plurality of types of the biological information so as to be in the descending order of the cullback / libra distance determined by the inter-distribution distance calculation;
An input order determining step for finally determining an input order based on the input order determined by the input order optimization;
A biometric authentication method comprising: an input biometric information instruction step for instructing the biometric information to be input in accordance with the input order finally determined by the input order determination.   8. The biometric authentication method according to claim 7, wherein the input order of the plurality of pieces of biometric information of the user determined initially is changed occasionally, and the changed input order is presented to the user. A program for executing authentication processing using a plurality of types of biometric information of a user obtained from a sensor by being executed by a CPU,
Biometric information matching means for calculating a matching score by comparing the biometric information of the user obtained from the sensor with the template of the user registered in advance in a database;
A person determination means for performing a person determination in an integrated manner using a sequential probability ratio test from the comparison scores for a plurality of types of the biological information obtained by the biological information collation means;
Distribution estimation means for obtaining, for each of a plurality of types of biometric information, a collation score distribution between the biometric information of the user (personal distribution) and a distribution of the collation scores of the biometric information of the user and the other person (distribution of other people). When,
An inter-distribution distance calculating means for obtaining a cullback / libra distance from the principal distribution and the other person distribution with respect to a plurality of types of biological information obtained by the distribution estimating means;
An input order optimizing unit that determines an input order of the plurality of types of the biological information so that the distance between the distributions is calculated in the descending order.
An input order determination means for finally determining an input order based on the input order determined by the input order optimization means;
A biometric authentication processing program that realizes input biometric information instruction means for instructing the biometric information to be input in accordance with the input order finally determined by the input order determination means.

Images