JPH06208611A - Device for authenticating individual person - Google Patents

Abstract

PURPOSE:To provide a device for authenticating individual person so as to eliminate dispersion such as a person easily getting identity confirmation permission, easily getting non-permission, easily mistaken for other person and hardly mistaken for other person. CONSTITUTION:The feature information of fingers previously extracted from the fingers of a lot of people is stored in a standard feature information storage part 12 as standard feature information. In the case of registering the feature information of the finger, a threshold value decision part 13 by individuals decides a threshold value for judging that person and the other person by individuals so as to provide a target second class error rate by collating the feature information to be registered with the respective pieces of standard feature information in the storage part 12 and stores the threshold value together with the feature information to be registered in a registering feature information storage part 14 and a threshold value storage part 15 by individuals. At the time of collation, the inputted feature information to be collated is collated with the correspondent registering feature information in the storage part 14 by a collation part 14. Corresponding to this collated result and the correspondent threshold value by individuals in the storage part 15, a judgement part 17 judges that person or the other person.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an entrance / exit management device for managing entrance / exit to an important facility, etc. Regarding the authentication device.

[0002]

2. Description of the Related Art In recent years, along with the development of an information-oriented society, there is a need for technology for identifying a person for the purpose of managing entry and exit from a specific area (important facility). In the past, personal identification was often done using a personal identification number or personal identification card (ID card), but these may be forgotten or lost, and there is also the risk of misuse by a third party. Yes, the limit of safety is becoming an issue.

Therefore, recently, a personal authentication method using the characteristic information of the human body (biological characteristic information) has attracted attention.
For example, various personal authentication methods using biometric feature information such as fingerprints, handprints, fingers, and retinas of eyes have been proposed.

Here, for example, a personal authentication device for performing personal authentication using the characteristic information of the finger will be described. The personal authentication device that uses the characteristic information included in the finger is a prism,
The finger image input means including a light source and a television camera takes in the image information of the finger, and the characteristic information of the finger obtained by adding the image information in the direction orthogonal to the finger is used for the matching.

When the abdominal surface of the finger is struck against the prism, the contact with the prism surface and the contact with the prism surface are not possible due to the unevenness of the finger skin formed by the fingerprint of the finger or the wrinkle pattern of the indirect part. . In this state, when the finger surface of the finger is illuminated by the light source through the prism, the light is totally reflected inside the prism in a portion where the skin is not in contact, and ideally all the input light enters the television camera.

However, the condition where total reflection is not satisfied at the portion where the finger skin is in contact, and light diffusion occurs,
The light incident on the television camera becomes part of the input light. As a result, the image information of the finger having a contrast corresponding to the unevenness of the skin is captured by the television camera.

Next, the characteristic information of the finger obtained by adding the image information in the direction orthogonal to the finger from the input image information of the finger is extracted, and this is registered (stored) in advance as personal characteristic information. Keep it. At the time of collation (at the time of individual authentication), the characteristic information of the finger obtained in the same manner is collated with the characteristic information of the registered finger to determine the person or other person.

The image information of the finger input at the time of collation cannot be input exactly the same as at the time of registration. Therefore, the characteristic information of the finger that is collated with the registered characteristic information does not completely match even if the information is obtained with the same finger.

Therefore, the degree of coincidence (matching value S) is calculated from both such feature information, and if this is smaller than the preset value, it is determined that both feature information are extracted from the same finger. I am trying to do it. That is, the person or the other person is determined as follows using an appropriate threshold value r. When S ≦ r: determined to be the person S> r: determined to be another person

Here, a method of determining the threshold value r will be described. First, m pieces of characteristic information for n persons are collected for each person. Note that n and m are, for example, n = 50.
0 and m = 5. When the collected n × m pieces of characteristic information are grouped into two groups and all the combinations are collated to obtain collation values, _nm C ₂ collation results (collation values) are obtained. Of these, the combination of this person is (
_m C ₂ ) × n. Here, it is organized as follows. Total number of combinations of this human T1 T1 = ( _m C ₂ ) × n Total number of combinations of other humans T2 T2 = _nm C ₂ −T1

Here, the obtained _nm C ₂ (= T1 + T2)
Consider that the judgment is performed with the threshold value r for each matching result (matching value). At this time, if the matching value is equal to or greater than the value r in the combination of T1 persons, it is an error that the person is mistaken for another person (first type error), and this number is e1 (r) pieces. Then, the first type error rate E1 (r)
Becomes E1 (r) = (e1 (r) / T1) × 100 (%).

On the other hand, in the combination of T2 other human beings,
If the collation value is smaller than the value of r, it is an error (second type error) in which the other person is mistaken for himself, and this number is e2.
If there are (r), the second type error rate E2 (r) is E2 (r) = (e2 (r) / T2) × 100 (%)

[0013] When r is changed within the range of possible matching values, e1 (r) and e2 (r) are functions of r, and are decreasing and increasing functions, respectively, so that E1
Similarly, (r) and E2 (r) are a decreasing function and an increasing function, respectively. It is ideal for a personal authentication device that both the first-class error rate and the second-class error rate be small,
In reality, the two have such a conflicting relationship, and it is difficult to make them compatible.

By the way, the personal identification device is often used for the purpose of ensuring security such as entry / exit control to an important area, and the second type error rate E2 (r) must be sufficiently small. So, for example, E2 (r)
R is selected so that 0.1%, and this is used as a threshold value for the determination of the person and others. When the value of r is determined, the first type error rate E1 (r) is also determined.

[0015]

As described above, in the conventional personal identification device, the judgment of the person or other person is
The second type error rate is set to a certain target value (for example, 0.1%) by using one threshold value common to all the registrants. Therefore, the second-class error rate referred to here means that it is an error rate as an average of the whole. The same applies to the first-class error rate.

By the way, as described above, when inputting with the same finger of the same person is performed twice, they do not completely match. That is, the characteristic information of the finger obtained for a plurality of inputs by the same finger has a certain range of variation. However, the range of such a variation varies from person to person, and if the person or other person is judged with a uniform threshold value for all registered persons,
For each individual, there was a problem that the error rate did not reach a certain level, and there were people who were likely to be permitted to confirm the identity and who were likely to be denied permission.

Therefore, the present invention provides a personal authentication device that eliminates variations among persons who are easily permitted to identify themselves, who are easily denied permission, who are easily mistaken for other persons, and who are not easily mistaken. To aim.

[0018]

The personal authentication device of the present invention extracts biometric information of an individual from biometric information input means for inputting biometric information of an individual and biometric information input by the biometric information input means. Feature extraction means, a first storage means for storing in advance biometric feature information of a plurality of individuals as standard feature information, and a biometric information input means and a feature extraction means when registering biometric feature information of an individual By collating the obtained biometric feature information with each standard feature information stored in the first storage means, a threshold value that is a predetermined false recognition rate of strangers is determined from the obtained plurality of collation results. The individual threshold value determining means and the individual threshold value determined by the individual threshold value determining means are used to determine the individual threshold value. Second storing in correspondence with the information
Collating the biometric feature information obtained through the biometric information inputting means and the feature extracting means with the corresponding biometric feature information stored in the second storage means when performing personal authentication. Means, and a judgment means for judging the person or other person based on the matching result of the matching means and the corresponding individual threshold value stored in the second storage means.

[0019]

[Function] The threshold for judging the person or others is
When registering biometric feature information of an individual in advance, by comparing this biometric feature information with a plurality of standard feature information stored in advance, the target second-class error rate (and first-class error rate) is obtained. By deciding on an individual basis, multiple registrants will each have the same second-class error rate, and some people are likely to get permission to confirm their identities, people who are apt to be denied permission, and people who are easily mistaken for others. , People who are not likely to be mistaken are eliminated.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a personal authentication device using, for example, characteristic information of a finger according to the present embodiment. In the figure, reference numeral 1 is an image information input unit (biological information input means) for inputting image information of a finger 2 to be collated.
A finger rest 5 comprising a finger rest 3 and a prism 4 embedded substantially in the center of the finger rest 3, and a light source 6 for illuminating the ventral surface of the finger 2 placed on the prism 4 of the finger rest 5. , A CCD type television camera (hereinafter abbreviated as TV camera) 7 for receiving the reflected light from the finger contact surface on the prism 4 and converting it into an electric signal. Reference numeral 8 denotes an A / D converter, which converts an image signal obtained from the TV camera 7 into, for example, 512 × 512 pixels and a quantization bit number 8
Digital image information IM (x, y) (0 ≤ x ≤ 511, 0 ≤ y ≤ 511).

Numeral 9 is a feature extraction for extracting the feature information P (x) of the finger 2 by adding the image information output from the A / D converter 8 in the orthogonal direction of the finger 2 as shown in Formula 1. Section (feature extraction means), and is connected to the control section 10. 2A shows an example of the image information IM (x, y) of the finger 2, and FIG. 2B shows an example of the characteristic information P (x) of the finger 2 extracted from this image information. Shows.

[0023]

[Equation 1]

Reference numeral 10 is a control unit that controls the overall control, and is mainly composed of, for example, a CPU (Central Processing Unit). Reference numeral 11 denotes an operation unit for inputting a personal identification number, which is composed of a keyboard or the like and is connected to the control unit 10.

Reference numeral 12 is a standard feature information storage section (first storage means) in which feature information of fingers extracted in advance from a large number of fingers is stored as standard feature information, for example, an optical disk device, a hard disk device, or a magnetic device. It is configured by a disk device or the like and is connected to the control unit 10. Reference numeral 13 denotes an individual threshold value determining unit (individual threshold value determining unit) that determines an individual threshold value for determining the person or other person by a process described in detail later. It is connected to the.

Reference numeral 14 denotes a registered characteristic information storage section (second storage means) for storing characteristic information of the finger of the registrant together with the personal identification number in advance, which is connected to the control section 10. 15
Is an individual threshold storage unit (second storage unit) that stores the individual threshold determined by the individual threshold determination unit 13.
Thus, it is connected to the control unit 10.

Reference numeral 16 is a collating unit (collating means) for collating the finger characteristic information input for collation with the finger characteristic information stored in the registered characteristic information storage unit 14, which is connected to the control unit 10. ing. Reference numeral 17 is a judgment unit (judgment means) for judging the person or other person based on the matching result of the matching unit 16 and the individual threshold value stored in the individual threshold storage unit 15, and is connected to the control unit 10. Has been done. Reference numeral 18 denotes an output unit for performing operation guidance and abnormality notification, which is composed of, for example, a lamp or a buzzer and is connected to the control unit 10. Next, the operation of the above configuration will be described.

The abdominal surface of the finger 2 is brought into contact with the prism 4 so as to include at least the first joint and the second indirect portion of the finger 2. The skin of the finger 2 is formed with irregularities due to fingerprints, streak patterns of indirect portions, etc. The convex portions contact the prism 4 and the concave portions do not contact. When the ventral surface of the finger 2 is illuminated from the light source 6 through the prism 4, the concave portion of the irregular surface of the finger 2 totally reflects the light and the convex portion diffuses the light. Therefore, the TV camera 7 can capture image information corresponding to the unevenness of the skin on the surface of the finger 2. The image information of the finger 2 thus obtained is digital image information IM (x, y) (0 ≤ x ≤ 511, 0 ≤ y in the A / D converter 8 having, for example, 512 × 512 pixels and a quantization bit number of 8. ≤ 511)

Is converted to Here, as for the image information of the finger 2, as shown in FIG. 2A, the positions of the TV camera 7 and the prism 4 are set in advance so that the longitudinal direction of the finger 2 is parallel to the x-axis. I shall. Next, in the feature extraction unit 9, the image information IM converted as described above.
From (x, y), the characteristic information P of the finger 2 is calculated by using the equation 1.
Extract (x). Next, a method of collating the characteristic information of the finger 2 will be described. Two pieces of feature information P1 (x) and P2
(X) is collated using Equations 2 and 3.

[0030]

[Equation 2]

[0031]

[Equation 3] Here, S is defined as a matching value between the characteristic information P1 (x) and P2 (x), and is expressed as follows for simplification. S = V (P1 (x), P2 (x))

It is considered that S is a kind of distance between the two pieces of feature information that take values from "0" to "1". The smaller the value of S, the more similar the pieces of feature information are. And the value of S is “0”. Generally, the threshold value r is used to determine the person or other person as follows. When S ≦ r: Judge as the person himself / herself When S> r: Judge as another person The present invention is most characterized in that the threshold value r is determined for each individual. It will be described in detail separately.

First, the creation of standard feature information will be described. Prior to registration of finger characteristic information, finger characteristic information extracted from a large number (for example, 10,000) of fingers in the standard characteristic information storage unit 12 (hereinafter referred to as standard characteristic information).
Need to be remembered. FIG. 3 is a flowchart showing the procedure for creating the standard feature information. Using the same method as described above, the finger feature information is set to 1 per person.
Individually collected and designated as PSi (x). PSi (x) (i = 1, 2, ..., N)

N is the number of registered finger feature information up to the present, and is incremented by one each time one person is registered. In addition, registration to standard feature information is limited to once per person,
Avoid duplicates by the same person over and over. Further, the person who registers in the standard feature information here does not have to be a registrant (user) who performs entrance / exit management or the like in the personal authentication apparatus.

The standard characteristic information may be collected by a personal authentication device for registering or collating the characteristic information of the fingers for entering and leaving the room, but it is generally for a large number of people, so it is generally used. Takes a huge amount of time. Therefore, in this work, it is realistic to collect in advance by another device having the same performance and copy the same standard characteristic information to the personal authentication device for registration and verification.

Therefore, in the present embodiment, the standard characteristic information storage unit 12 is composed of, for example, a portable optical disk device, etc., and a copy of the standard characteristic information collected and created by another device is made. It is stored in an optical disc and set in a personal authentication device for registration and verification for use.

Next, the registration of the characteristic information of the finger will be described. This is the registration of the finger feature information of the registrant (user) who is managed to enter and leave the room by using the personal authentication device, and FIG. 4 is a flowchart showing the procedure for registering the finger feature information.

The registrant has been previously given a serial number j as a personal identification number by the system administrator. In the registration work, first, the serial number j is input from the operation unit 11. Then, the finger 2 is placed on the prism 4, the image information of the finger 2 is input, and the characteristic information P of the finger 2 is input.
It is assumed that j (x) has been obtained. The individual threshold value determining unit 13 collates the registered characteristic information Pj (x) with all of the standard characteristic information PSi (x) in the standard characteristic information storage unit 12, and compares the collation results, that is, n. The individual matching value Si is obtained. Si = V (Pj (x), PSi (x)) (i = 1,2, ..., n) Next, the n matching values (S1, S2, S3, ..., Sn) obtained here. Are sorted in ascending order, and are (SS1, SS2, SS3, ..., SSn). Then, the number (n × 0.0) corresponding to the target second-type error rate E2 (for example, 0.1%) is selected from these.
01) is selected in ascending order of matching value. k = [n × E2 / 100] ([] is Gaussian symbol)

At this time, from the smallest to the kth (SS
1, SS2, SS3, ..., SSk) are the targets for causing the type 2 error for the registration characteristic information Pj (x), and thus the maximum value, that is, S
If Sk is selected as the individual threshold rj = SSk of this registrant, the second-class error rate of this registrant will be E2 (%) as desired.

The threshold value rj thus determined is stored in the individual threshold value storage section 15 and the registered characteristic information storage section 14 together with the registered characteristic information Pj (x) under the control of the control section 10.

Next, the collation process (personal authentication process) will be described with reference to the flowchart shown in FIG. First, the collated person inputs the serial number j as the given personal identification number from the operation unit 11. The control unit 10 receives the input serial number j, and registers the registered characteristic information Pj (x) and the individual threshold value rj stored in the corresponding area, respectively. It is read from the threshold value storage unit 15.

On the other hand, the person to be collated receives an instruction to place the finger 2 on the prism 4 by the lamp of the output unit 18 or the like, and the characteristic information P (x) of the finger 2 is extracted. In the control unit 10, the registered characteristic information Pj (x) read from the registered characteristic information storage unit 14 and the extracted collated characteristic information P
(X) is sent to the collation unit 16, and collation processing is performed by the following equation. S = V (P (x), Pj (x))

Then, the control unit 10 sends this collation result (collation value S) to the determination unit 17 together with the individual threshold value rj read from the individual threshold storage unit 15 to use the following equation. To determine the person or other person. When S ≦ rj: Judge as the person himself When S> rj: Judge as another person

As described above, according to the above embodiment,
When registering the characteristic information of the finger in advance, the threshold value for determining the person or other person is collated with the standard characteristic information of a large number of people who has collected and stored the characteristic information in advance, so that the target Since the second-class error rate (and the first-class error rate) is determined for each individual, a plurality of registrants each have the same second-class error rate. Therefore, there is no variation among people who are easy to pass, people who are difficult to pass, people who are easily mistaken for other people, people who are hard to be mistaken for others.

In the above embodiment, the personal authentication device that uses the signal obtained by adding the image information in the direction orthogonal to the finger from the image information of the finger as the characteristic information of the finger has been described as an example. The invention is not limited to this, and can be similarly applied to a personal authentication device that uses characteristic information such as a fingerprint, a handprint, and a retina of an eye.

[0046]

As described above in detail, according to the present invention, there is no variation among persons who are likely to be permitted to identify themselves, who are likely to be denied permission, who are easily mistaken for others, and who are not easily mistaken. A personal authentication device can be provided.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a personal authentication device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of finger image information and finger characteristic information.

FIG. 3 is a flowchart showing a procedure for creating standard characteristic information.

FIG. 4 is a flowchart showing a procedure for registering finger feature information.

FIG. 5 is a flowchart showing a procedure for explaining a matching process.

[Explanation of symbols]

1 ... Image information input section (biological information input means), 8 ... A
/ D converter, 9 ... Feature extraction unit (feature extraction means), 10
...... Control unit, 11 ...... Operation unit, 12 ...... Standard feature information storage unit (first storage unit), 13 ...... Personal threshold value determination unit (Personal threshold value determination unit), 14 ...... Registered characteristic information storage unit (second storage unit), 15 ... Individual threshold storage unit (second storage unit), 16 ... Collation unit (collation unit), 17 ... Determination unit (determination unit) .

Claims (1)

[Claims] 1. Biometric information input means for inputting biometric information of an individual, feature extraction means for extracting biometric characteristic information of an individual from biometric information input by the biometric information input means, and biometric characteristics of a plurality of individuals in advance. First storage means for storing information as standard feature information, and biometric feature information obtained through the biometric information input means and feature extraction means when registering biometric feature information of an individual is stored in the first storage means. Individual threshold value determining means for determining a threshold value that is a predetermined false recognition rate from a plurality of matching results obtained by matching each standard feature information stored in Second storage for storing the individual threshold value determined by the threshold value determining means in association with the biometric feature information obtained from the feature extracting means used in determining the individual threshold value And a verification means for verifying the biometric characteristic information obtained through the biometric information input means and the feature extraction means with the corresponding biometric characteristic information stored in the second storage means when performing personal authentication. A personal authentication device comprising: a judgment means for judging the person or other person based on the matching result of the matching means and the corresponding individual threshold value stored in the second storage means.