JP2001212112A - System and method for recognizing and identifying subject under a series of changes controlled by the subject to biometrics, and record medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system and method for recognizing and identifying a subject under a series of changes controlled by the subject to biometrics and a record medium. SOLUTION: This invention defines new biometrics called as resultant biometrics. The resultant biometrics are a combination of the conventional biometrics and a change specifically controlled. In other words, signals of the resultant biometrics are arranged as a sequence of a short-interval biometrics signal and altered according to a specified pattern. A finger print or a palm print thus obtained is turned to a continuous pressing image, for example, by letting a subject apply a torque or a rotation or both thereof for an image gaining time length (a controlled change). A physical method of letting the subject deform the image makes an action apart of the resultant biometrics and the finger print of the palm print a physiological part of the resultant biometrics.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

FIELD OF THE INVENTION The present invention relates to the field of biometrics, that is, the physiological or behavioral characterization of a subject that is more or less uniquely associated with the identification of the subject. More particularly, the present invention relates to a novel type of biometrics formed by a controlled series of changes over conventional biometrics.

[0002] The present application is a provisional application filed on December 2, 1999, application number 60 / 168,540 (IBM).
Docket No. YO999-588).

[0003]

BACKGROUND OF THE INVENTION Fingerprints have been used for at least 50 years.
Used for decades in semi-automatic identification of individuals for the purpose of imparting legal enforcement and in automatic authentication applications to control access such as building access and computer login. Have been. Signature recognition for automatically authenticating personal identification has been used for at least 15 years, mainly for banking applications. The first stage in an automatic fingerprint identification system or an automatic signature identification system is a signal acquisition stage for acquiring a fingerprint or signature of an object. Some techniques for obtaining fingerprints include:
Scanning ink-based or ink-free fingerprints using optical, capacitive, and other semiconductor-based detection techniques can be mentioned. The acquired signal is processed and matched against a stored template that is a mechanical representation of the fingerprint. Image processing techniques typically align the projections and valleys of a fingerprint and derive a template from the saliency and valley patterns of the fingerprint image.

[0004] Signatures, on the other hand, are detected using a pressure-sensitive writing pad, or an electro-magnetic writing device. In more advanced systems,
Use a special pen to calculate pen speed and acceleration. The recorded signal can be a list of simple (x, y) coordinates for static signature recognition, or the coordinates can be (x (t), y
(T)). The template representing the signature is more directly related to the acquired signal than the fingerprint template. For example, a signature can be represented as a set of strokes in which acceleration is coded for each stroke between a plurality of ends. Examples of signature authentication include: S. N
alwa, "On-line automatic signature certification", Proceedings of IEEE, pp. 215-239, 1
February 997 can be mentioned.

[0005] Biometrics such as fingerprints, signatures, faces and voices have recently been used, for example, for access to medical
Access to M, access to Internet services, and other similar applications are increasingly being used to authenticate user identities.

[0006] With the rapid progress of the Internet, many e-commerce and e-business applications have been developed and used. For example, retail purchases and travel reservations using credit cards over the Internet are very common commercial applications. Today, for identification and authentication, respectively, a user is identified by a user ID and password. fingerprint,
More robust methods for authentication and possibly identification, including biometrics such as signatures, voiceprints, iris images, and facial images, replace the simple methods described above for identification. That will be very close. Automated biometrics
The system includes obtaining a signal from the user that more or less uniquely identifies the user. For example, for fingerprint-based authentication, the user's fingerprint is scanned and some representations need to be calculated and stored. Authentication is then performed by comparing the expression extracted from the user's newly acquired fingerprint image with the stored expression extracted from the image stored at the time of registration. In a speaker verification system, a user's speech signal is recorded, and several expressions are calculated and recorded. Authentication is performed by comparing the recorded speech signal at the time of access or logon with the recorded representation. Similarly, for a signature certificate, the template is extracted from the digitized signature and compared with a pre-calculated template.

[0007] Biometrics are broadly divided into two groups: behavioral biometrics and physiological biometrics. Physiological biometrics are relatively constant biometrics over time, such as fingerprints, irises, and the like. Behavioral biometrics, on the other hand, can change slowly over time or suddenly in a shorter time. These biometrics can include, for example, signatures, voices, and faces. (Faces are often considered physiological biometrics because their basic characteristics do not change easily, but aging, hair conditioning, The growth of the beard and facial expressions are to change the expression of the face as a whole.) The field of the invention relates to physiological and behavioral biometrics, and more particularly, the present invention can be used as a novel type of biometrics, a series of user controlled behavioral changes It relates to physiological or behavioral biometrics, such as changes.

[0008] Commerce / business solutions based on the Internet are accessible from remote, unmanaged locations, such as a user's home. For this reason, the biometric signal needs to be obtained from a remote user who is not supervised. For this purpose, a fingerprint or palm print reader, a signature digitizer or a camera for obtaining a face or iris is attached to a computer at the user's home. This, of course, gives the possibility of unauthorized and unauthorized system access. By intercepting, a malicious individual or institution can obtain a biometric signal from the true user through the network or obtain the signal from an application in which the user is using his or her own biometrics. Become. The recorded signal will then be reused for unknown and unauthorized purposes by acting as a registered genuine user of the Internet service. The simplest way is to acquire the signal once and reuse it several times. The signal previously acquired can be slightly disturbed, producing a biometric signal that looks "fresh". If a perfect fingerprint or palm print is known to the intruder, a more advanced method is to produce a three-dimensional copy of the artificial (fake) finger or palm from a material such as silicon or latex. is there. The genuine user's fingerprint or palm print image would then be produced by the impostor without much effort. The transaction server, authentication server, or other computing device may be configured or obtained such that the biometrics signal communicated from the client is either a raw signal at the time or not previously acquired. It will play a role of guaranteeing that the signal is obtained. With the use of artificial body parts, many fingerprints and palm print readers are considered very legitimate to outsiders if the proper materials are used to manufacture these body parts. Give the image to be. These images are often also considered authentic to the processing components that are part of the authentication system. Thus, it is extremely difficult to determine whether a static fingerprint or palm print image was provided by an authentic finger or palm, or by a fake copy. Another physiological biometric has similar limitations, in which the iris of the true user will be imaged and used for unauthorized access. Good iris recognition systems detect fast oscillations of the iris diameter, but this phenomenon can also be mimicked by a sequence of iris images. A similar method can, of course, be used for facial biometrics.

[0009]

Fingerprints and, to a lesser extent, palm prints, access to medical documents, ATM
Access, and other similar applications, are increasingly being used to authenticate user identities. The problem with the conventional identification method is that a three-dimensional fake fingerprint or palm print can be produced. Silicone, latex, urethane, and other materials can be used to manufacture these artificial body parts, and most image acquisition devices are difficult to distinguish from artificial stamps It easily gives a stamp that is considered authentic on the protruding part. The cause is that the given fingerprint or palm print is a static drawing of the image at a certain moment. The fingerprint is
It is not time dependent. Similar problems exist for biometrics such as faces and irises.
The problem in this case is that fake copies of two or three dimensions are produced and biometrics security
It is used to deceive the system, because the biometrics are not time dependent.

[0010] Another problem in the prior art is that during acquisition of the biometric signal, only a single stationary fingerprint or palm print is collected. This instantaneous image is created by the user applying force or torque to the image acquisition device (fingerprint or palmprint reader) with his or her finger,
The projections and valleys of the fingers or palm may be deformed. The problem is that one or more images are collected, ie, it is not possible to detect whether the images were acquired without deformation without changing the mechanism of the sensor. In addition, the prior art has the problem that only one image can be selected for use in identifying an individual. Of course, for non-contact biometrics such as the face and iris, deformation of the biometric pattern is either undetectable or very difficult to detect.

The following documents describe the prior art. Aalen Pu and Demetry Psaltis, "User identification by continuous fingerprint input", US Patent No. 5,933,515 The method disclosed by Pu and Psaltis in US Patent No. 5,933,515 Are used and a large number of fingers are used in a sequence known only to the user. If the fingers are numbered from 0 to 9 from left to right, for example, the sequence is nothing but a PIN. If one considers the sequence described above plus a fingerprint image as one biometric, the sequence can be modified and becomes a non-stationary part of the biometric. However, this technique is not a series of controlled changes to existing biometrics, and the pattern of each finger remains the same,
The sequence pattern changes. The problem is that because fingerprint entry is a slower process, anyone can see the sequence of fingerprints more easily than looking at PIN entries. Furthermore, it is necessary to store each fingerprint of the subject for comparison.

Another problem with the prior art is that sensors (fingerprint or palmprint readers, facial expression cameras) are required to verify the authenticity of the biometric signal, for body part authentication or for sensor authentication. The need to incorporate computational resources. Body part authentication is commonly performed by heart rate and body temperature measurements. The sensor authentication is performed by two-way communication between the sensor and the authentication server in the form of a question and answer session.

One major problem with conventional palm prints and fingerprints is that if the user loses the fingerprint or palm print for any reason, or if the lost hand has a template representation of the image, replacing the image is not an option. The inability to do so results in the image being permanently uncertain. For this purpose, another finger image can be used, but only a few times.

A problem with conventional systems that use stationary fingerprints is that there is no information that can be used with the fingerprint and that can be used in addition to its discriminating ability. That is, individuals with apparently close fingerprints will be confused because the fingerprints are static and no additional information is available to distinguish these images.

In the conventional fingerprint database, a search is performed by first filtering the fingerprint type (hoofprint, vortex,...). The problem with this prior art is that the fingerprint image is a temporally static snapshot and there is only a small number of fingerprint classes, since no additional information accompanies the fingerprint.

The final problem with any prior art biometric is that no backward comparison is made with other biometrics. In other words, the use of the face for authentication is not compared backwards with the fingerprint database.

It is an object of the present invention to provide a novel type of biometrics provided by a subject through a series of controlled changes in existing biometrics.

Another object of the present invention is a biometric that is modified through a series of user-controlled changes having both physiological and temporal properties.

Another object of the present invention is that it is modified through a series of user controlled changes that include physiological, physical (eg, force, torque, linear motion, rotation), or temporal properties and both. Biometrics.

It is an object of the present invention to provide an efficient method for modifying unreliable biometrics.

[0021] It is an object of the present invention to provide a biometric that is modified through a series of user controlled changes that combine a user selected behavioral biometric with a conventional biometric.

A further object of the present invention is a biometric that is more difficult to provide with a forged body part.

[0023]

SUMMARY OF THE INVENTION The present invention achieves the above or other objects by employing a novel class of biometrics that is modified through user-controlled changes, called result biometrics. Things. Biometrics refers to more or less unique properties associated with an individual. The biometrics include physiological biometrics and behavioral biometrics. Physiological biometrics are properties that do not change or change very slightly over time, while behavioral biometrics change over time because they depend on an individual's atmosphere, mental or physical state, and suddenly change. This is a characteristic that may change to Examples of physiological biometrics include fingerprints, irises, and faces, and examples of behavioral biometrics include voices and signatures.

The biometrics introduced in the present invention may use behavioral biometrics or physical elements or both to alter the appearance of physiological or behavioral biometrics (by controlling changes over time). ) Results in a combination of physiological or behavioral biometric signals provided by the subject. In a preferred embodiment, the resulting fingerprints and palm prints force the appearance of the image as the image is scanned,
It can be changed by changing the torque and rotation. This allows the fingerprint or palm print image to be
That is, the imprints of these images are continuous and elastically deformed according to forces, torques, rotations, or physical elements, or behavioral biometrics (eg, signatures or gestures), or both, and Is done.

[0025]

DETAILED DESCRIPTION OF THE INVENTION The present invention introduces a new biometric called result biometric. Resulting biometrics is a method of converting physiological or behavioral biometrics over a predetermined time interval, adding a second behavioral biometric to a first biometric signal recorded at a predetermined sample rate. Resulting in a continuous sequence of physiological or behavioral biometric signals. This transformation is obtained as a series of user controlled changes to the first biometric.

Biometrics, such as conventional fingerprints, have been used for (automatic) authentication and identification for decades. The signature is recognized as legally binding proof of identity and automatic signature verification /
The certification method has been available for at least 20 years. FIG. 1 is a diagram illustrating these biometrics. The signature 110 is shown at the upper left, and the fingerprint imprint 130 is shown at the upper right. The voice (voiceprint) is shown in the lower left, and the iris pattern is shown in the lower right.

Biometrics can be used for automated authentication or identification of a subject (human). Typically, the subject is remembered by providing sample biometrics when opening a bank account, registering for an Internet service, and the like.
From this sample biometric, a template is stored and used for matching purposes at the time the user wishes to access the account or service. In the preferred embodiment described, the template for the resulting biometric is a combination of a conventional template for the biometric and a template that records the apparent change in the biometric over time.

The obtained fingerprint and palm print will be described later in more detail. The fingerprint or palm print template is derived from the selected imprint in a sequence of no force, torque or rotation. The trajectory template quantitatively describes the motion trajectory of the obtained fingerprint with respect to time. Matching two templates is the opposite of combining a traditional fingerprint template with dynamic string matching of a trajectory similar to signature matching. This string matching is well known in the prior art. The fingerprint obtained by the user detecting while applying the torque will be described in further detail.

Biometrics identify more or less individuals, and a given biometric signal is such that only one individual owns the biometrics or significantly narrows the list of people who own the biometrics. Fingerprints are an excellent biometric, as no two individuals with the same fingerprint have been historically found so far. On the other hand, biometric signals such as shoe size and weight are not sufficient as biometric signals, as it is clear that these signals have little screening value. Biometrics can be distinguished between behavioral biometrics and physiological biometrics. Behavioral biometrics are variables that depend on the physical and mental state of an individual and can change rapidly over time. Behavioral biometrics
A signature 110 and a voiceprint 120 can be cited (see FIG. 1). Physiological biometrics, on the other hand, are less variable. The fingerprint 130 shown in FIG.
The basic flow structure of the ridges and valleys shown in FIG.
Other examples of biometrics include the circular texture of the subject's iris 140 shown in FIG.
It is considered that there is little change in this over the span of the subject's life. Thus, for example, 11
For existing behavioral biometrics such as 0 and 120, existing physiological biometrics (iris 140) that can be controlled to some extent by the subject and whose appearance is not affected, or are only slightly affected (Fingerprint 130) Physiological biometrics are present. The signature and voiceprint on the left are behavioral biometrics; the fingerprint and iris image on the right are physiological biometrics.

Here, reference is made to FIG. A typical conventional automated fingerprint authentication system uses a fingerprint image (biometric signal) as an input 210 to a biometric matching system. The system comprises three other steps 215, 220, 225, which are a signal processing step 215 for extracting features, a step 220 for extracting templates from features, and matching of templates. 225. Along with the biometric signal 210, the subject's identification unit 212 is input to the matching system. Template matching stage 225
In the meantime, the template associated with this particular identifier is
The template indexed by the identity (identification unit) is retrieved from a predetermined database 230.
Extracted template 220 and database 230
If a match / no match is made between the template and the retrieved template, a “Yes / No” response 240 is output from the matching system. Matching is typically performed by a measure of similarity; if this measure is significantly higher, the response is "Yes"; otherwise, the response is "No". The following references are examples describing the prior art. N. K. Lhasa, S.M. Chen, A. K. Jane, Adaptive Flow Orientation Based on Feature Extraction of Fingerprint Images, Pattern Recognition, Vol. 28, No. 11, 1657-
1672 pages, November 1995

The system 200 is not limited to fingerprint authentication, and the architecture of the system is appropriate for any biometric. The biometric signal 210 input to the system is either local to the client's application or remote, running a matching application at a given server. Thus, the architecture 200 can be applied to all biometrics and networked or non-networked applications.

The system 200 shown in FIG. 2 is an authentication system, while the system 250 shown in FIG. 3 is an identification system. Typically, in conventional flow-based automated biometric signal identification systems, input 210
(See FIG. 2) only. Again, the system is composed of three stages 215, 220, 225,
These steps are signal processing steps 215 for feature extraction.
And step 220 of extracting a template from features and a template matching step 225. However, in the identification system 250, the biometric signal 210 is input to the system. During the template matching stage 225, the extracted templates are
A match is made with the pair of identifiers recorded in 30. If there is a match between the extracted template 220 and the template containing the identity in the database,
The identity is used as the output 225 of the identification system 250. If there is no match in the database 230, the output identity 225 is set to NIL.
Again, the biometric signal 210 is either localized in the client's application or is an application that is remotely matched on a given server. Thus, the architecture 250 can be applied to networked or non-networked applications.

The biometric signals are combined (integrated) at the system level and at the subject level.
Integration at the subject level is an object of the present invention. The system is summarized in FIGS. 4-7 for comparison with different methods and to show how to determine biometrics at the integrated subject level (result biometrics). Four possibilities for combining (integrating) the two biometrics are shown in FIGS. Performing AND210 (FIG. 4), O
Performing the binding with R220 (FIG. 5), ADD230
(FIG. 6) and combining with the sequential 240 (FIG. 7). 2 of subject Z
One of biometrics _B x (250) and _B y (2
60) is used for authentication as shown in FIG. However, two or more biometrics of a subject can be combined in a straightforward manner. These biometrics can be the same, for example, two fingerprints, or different biometrics, for example, a fingerprint and a signature. Matching means, matching means A20 of FIG. 4 respectively corresponding for biometrics _{B x} and _{B y}
2 and matching means B204. These matching means compare the input biometrics 250 and biometrics 260 templates with the stored templates and compare them with “Yes / No” 21
4 response is provided in system 210 and system 220 or system 230 and system 240
As in, the score values S ₁ (231) and S
₂ (233) is given.

The system 210 combines the two “Yes / No” responses of the matching means A 202 and the matching means B 204 by performing an AND, and combines the results through an AND gate 212. Thus, only if both matching means 202 and 204 match, the "Yes / No" response 216 of system 210 will be "Yes" (both biometrics match and subject Z is authenticated). Otherwise, the output 216 is "No" (one or both biometrics do not match and subject Z is rejected). The system 220 performs combining through OR, and includes a matching unit A202 and a matching unit B.
The results are combined by OR gate 222 using the two “Yes / No” responses at 204. Therefore, if the “Yes / No” output 214 of one of the matching means A 202 and the matching means B 204 is “Yes”, the “Yes / No” output 216 of the system 220 is:
"Yes" (one or both of the biometrics match and subject Z is authenticated). "Yes / N" of matching means 202 and matching means 204
o "only if both outputs 214 are" No "
The “Yes / No” output 216 of the system 220 indicates “N
o ”(both biometrics do not match,
Subject Z is excluded).

For system 230,
Matching means A 202 and matching means B 204
Combined, matching score S₁(231) and
S₂(233) are generated. Score S₁
Is Biometrics B_xExtracted from (250)
To what extent the template is recorded in the matching means A202
Indicates whether the template is similar to the template
Score S₂Is Biometrics B_yFrom (260)
To what extent the extracted template matches the matching means B
Similar to the template stored in 204
It represents what. The addition means ADDer232 is
Sum of cores 231 and 233, S₁+ S₂Is output. 2
At 34 this sum is compared with a decision threshold T and S₁+ S
₂> T (236), it is determined as “Yes”,
Biometrics B_xAnd B _yThe subject Z having
Otherwise, the output is "No" (238)
The subject is excluded.

The system 240 of FIG. 7 is bonded continuously biometrics subject Z and (250) and _B y (260). First, Biometrics B _x (250)
Is matched with the template stored in the matching means A (202), and the matching score S ₁ (23
Obtain 1). The obtained matching score is compared with a threshold value T ₁ (244). If the test does not pass the test 244, the output is set to “No” (238) and the subject Z is excluded. Otherwise, Biometrics B
_y (260) is again matched with the template stored in the matching means B (204). The output score S ₂ (233) of this matching means is equal to the threshold T ₂
(246). Output "Yes", that is, if _{S 2> T} 2, subject Z is authenticated. Otherwise, if the output is “No” (238), the subject Z is excluded.

FIG. 8 is a general flow diagram for combining user actions, ie, combining biometrics at the subject level. The user's action is a second behavioral biometric, just like the movement of the pen when signing. User 410 provides conventional biometrics 420 for authentication or identification purposes. Such biometrics can include a fingerprint, iris, or face. However, rather than keeping the biometrics stationary as in the case of fingerprints or faces, or keeping the eyes open as in the case of iris recognition, Specific operation 430, α
Execute (t). This operation is performed for a time 432 from time 0 (434) to a predetermined time T (436). Thus, the action α (t) is a predetermined one-way function for time 430 and acts as conventional biometrics 420. This biometric is
It is the actual biometric of the user 410 and not a machine-readable signal (ie, in the case of a fingerprint, a three-dimensional finger with a pattern on the finger). Although constraints on operation 430 are defined, within this constraint the user 410 can define the operation (e.g., the constraint on signing is that the user places the pen on the paper in the x, y directions). But the pen cannot be moved in the z-direction). That is, the predetermined operation 430 converts the user's biometric with respect to time. The input to the biometric signal reading device 460 is this converted biometrics 450. Output 470 of this device
Between time 0 (434) and time T (436)
Individually converted biometrics signal B (t) 4
There are 80 sequences (series). In the case of a fingerprint, a plurality of fingerprint images are obtained, and in the case of a face, a plurality of face images are obtained. This output sequence 470 is used as an input 485 to a predetermined extraction algorithm 490. The extraction algorithm consists of a pair of transformed biometrics (α ′ (t),
B) Perform calculations from the sequence of 495. Function α ′
(T) is a predetermined behavioral pattern over the time interval [0, T] of biometrics B and is related to the function α (t) selected by the user (very similar to the user selecting a signature). . Biometrics B is calculated from the pair (α ′ (t), B), and α (t) 430 is
Zero is when the user has taken no action,
The output 470 is the undeformed digitized biometrics 420. In general, the biometrics 420 in the signal 480 can be calculated without deformation.

Here, reference is made to FIG. FIG. 9 is a diagram illustrating the resulting biometrics of a fingerprint (without sliding against a glass plate) with the user allowed to rotate the finger on the fingerprint reader 510. Such rotation is performed at a predetermined angle α defined by the user according to a time function α (t). An example of the resulting fingerprint is shown in FIG. The user places finger 540 on fingerprint reader 510 at hand position 520. Thereafter, from time 0 (434) to time T (436), the user rotates the finger 540 on the glass plate according to a predetermined angle α with a function of time α (t). This rotation takes place in a horizontal plane,
This plane is parallel to the glass plate of the fingerprint reader. The function of rotation in this case is the behavioral part of the resulting fingerprint, which is defined by the user (if this part of the resulting biometrics is uncertain, the user will replace this behavioral part of the resulting fingerprint with Can be specified again). First, the user rotates left by an angle 550 to a hand position 525. Thereafter, the user rotates right by an angle 555 to a final hand position 530. During this operation over the time interval [0, T], the fingerprint reader produces 470 the transformed (deformed) fingerprint image as a series. This output 470 is the transformed biometrics 480 as shown in FIG.
(Fingerprint), and these converted biometrics 480 are input to the extraction algorithm 490 (FIG. 8). This algorithm calculates the angle α from the time 0 (434) to the time T (436) using the given output 470 as a function of time α (t).
The obtained fingerprint is (α (t), F) in this case, where F is an undeformed fingerprint image. An undeformed fingerprint image is found at times 434, 570, 436, where the rotation angle α is zero.
A preferred method for extracting the rotation angle from the deformed fingerprint image is shown in FIGS.

FIG. 10 is a diagram showing an example of fingerprint biometrics obtained when the user has four degrees of freedom instead of the one degree of freedom shown in FIG. 9 for moving a finger on the scanner. Again, as shown in FIG.
The finger 540 can be rotated about an axis, which is perpendicular to the glass plate of the fingerprint reader 510. This is 55
It is indicated by a rotation α to the left along 0 and then to the right along 55. This is done by bringing the hand from 520, 525 to the final hand position 530. Also, at any given angle α, the user can perform rotation β 610 about finger axis 620. Finally, the user applies the force f630 to the fingerprint reader 510.
Add in parallel to the glass plate. This force is
It can be concentrated only in two directions or not as 634. In the former case, which concentrates in the direction of the finger 632, the degree of freedom (without sliding the glass plate) for moving the finger is 3, and in the latter case, in which the force is not concentrated, The degree of freedom is four. Applying a force f to the angles α, β is combined and is referred to as the movement m. During the user's operation over the time interval [0, T], the fingerprint reader outputs 470 a sequence of transformed (deformed) fingerprint images. This output 470 is a sequence 480 of transformed biometrics (fingerprints) as shown in FIG.
0 is input (FIG. 8). This algorithm compares the given output 470, angle α, and angle β with the interval 0
(434) Calculated as a function of time over time T. For example, α (t), which is a function of the angle α, is the function 560 in FIG. Furthermore, the algorithm uses the force 630
Is calculated. When the force is concentrated along the finger direction 632, f (t) is a one-way function. If a force is applied along the glass plate of reader 510 in either direction, f (t) will be a two-way function. The force thus has a component in the x direction and a component in the y direction. The fingerprint obtained in this case is (α (t), β
(T), f (t), F) or (m (t), F), where F is an undeformed fingerprint image. The undeformed fingerprint image has the reconstructed motion m
Times 434,570,4 when (690) is zero
Found at 36.

FIG. 11 is a diagram showing an embodiment in which the same principle as that of a fingerprint is applied to a palm print. The palm print reader 710 includes a glass plate 720, which can be mounted, for example, next to a door. Only the user matching the palm print template is authenticated and access is permitted. The user places his hand 730 on the palm print reader plate 720. As with the resulting fingerprints shown in FIGS. 9 and 10, the user does not need to keep the palm biometrics stationary, but exercises the palm. In the case shown in FIG. 11, rotation of the palm about an axis perpendicular to the glass plate is a behavioral part of the resulting palm print biometrics. The user may, for example, put his hand on the right 740
, Then rotate the hand to the left 744, and then rotate the hand again to the right 748. As shown in FIG. 9, during these operations over a predetermined time interval [0, T], the palm print reader outputs a sequence of transformed (transformed) palm print images. This output is a sequence of biometrics (palmprint) converted as shown in FIG. 8, and these biometrics are input to the extraction algorithm 490 as shown in FIG. The algorithm calculates a given palmprint reader 710 output and a palm rotation angle α as a function of time α (t) from time interval 0 (434) to time T (436). The obtained palm print is (α (t), P), where P is an undeformed palm print image. An undeformed palm print image has a time when the rotation angle is zero.
34,570,436.

FIG. 12 is a diagram showing biometrics obtained using a face. The subject 800 is posing in front of the camera for recognition or identification of both the biometrics of the physiological face plus the behavioral part. This behavioral part is introduced by movement of the subject's head. This movement is a function of time Face-i
Generate a sequence of facial images as mage (t). If the subject's face is at the canonical position, the head is embedded in a coordinate system 805 with the Y-axis 810 along the length of the head. X-axis 820 is parallel to the line connecting the ears, and Z-axis 830 is perpendicular to the plane in front of the face. In this case, the subject converts the obtained biometrics Face-Image (t) into Y
Pan840 is generated as a function of time, Pan (t), generated by shifting the face around the axis.
Get. The subject further tilts the face by bending the face in a plane 860 that spreads in the Y-axis and the left and right directions (85
0) You can.

This slope 850 is a function of another time Til
gives t (t). Therefore, the biometrics obtained in this case are the face image Face-Image (t ₀ ) at the predetermined time t ₀ , the front image of the face, the left and right, the inclination Pan (t), and the Tilt (t). The facial image in the sequence is a mathematical transformation of the facial image. In the present invention, other variations of the face image through other means can be used.

FIG. 13 shows a block diagram 900 of a general process for extracting behavioral portions from the resulting biometrics. Input 910 is a sequence of B (t) of the biometric signal. At block 920, the following biometric signal B (t + 1)
And B (t) are processed through inter-signal analysis. In block 930, the change α (t in the action portion
Use this analysis to extract +1) -α (t). This also gives the output a (t) as a function of time 940, where a (t) is the resulting biometrics B
This is the action part of (t). In FIG. 13, characteristic steps (inter-image flow analysis and affine motion parameter estimation) have been added to estimate finger motion from the sequence of deformed fingerprint images generated in FIG. These are further described in FIG. 14 and FIG.
5 will be described in detail.

The rotation from one fingerprint image to the next can be estimated using the steps described in FIG. Images B (t) and B (t + 1) 1010,1
015 is a 16 × 16 block 1020, 1022,
1024,. . . Into 1028 as given as the MPEG compression standard. Given a fingerprint image sequence B (t), which is the two images (1010 and 1015) shown in FIG. 14, the inter-image flow for each block (size is 16 × 16) 1030 present in the image. (U, v) 10
40 is calculated. This is for any image B (t)
1010 also shows the motion expressed for the image B (t + 1) 1015 immediately after in the sequence. Flow characterization [u (x, y), v (x,
y)] 1050 is a function of (x, y) 1060 and t 1070 of the image sequence, and is thus a constant motion representation for consistent interpretation. The motion expression 1050 is MPEG-1 or MPEG-
It can be calculated from the coarse motion vectors coded in the two image sequence. If the input is uncompressed, the flow field can be estimated using motion estimation techniques known in the art.

The following references respectively disclose the current technology of MPEG compression, an example of optical flow estimation within an image sequence, and a conventional example of extracting a flow directly from an MPEG-compressed image sequence. Things. B. G. FIG. Haskell, A.S. Pulley, A. N.
Netrabi, Digital Video: An Introduction to MPEG-2, Chapman and Hill, 19
1997, J.M. Bergen, P.S. Anandan, K .; Hannah,
R. Hingorani, hierarchical model-based motion estimation,
Second European Conference on Computer Vision, 237-252, 199
2 years, Citra Doray and Vikrant Cobra, H
Extraction of Motion Annotations from MPEG-2 Compressed Video for DTV Content Management Applications, IEEE International Conference on Multimedia Computing and Systems, pp. 673-67
8, p. 1999.

Here, reference is made to FIG. Block 10
20, 1022, 1024,. . . , 1028, the largest connected portion of the zero-motion block, which is shown as a pivot region in FIG. 15, is determined by examining the flow [u (x, y), v (x, y)] 1050. . Further, the flow is analyzed in this region. Using the flow calculated for each image in a given image sequence, the motion parameters from the fingerprint region impose an affine motion model of the image-image flow, radially around a bounding box 1120 of the region 1110. Can be calculated by sampling the non-zero motion block. The affine motion A1130 has a shape 1140
To the shape 1145 shown in FIG. 16 and the translation 1150, rotation 1152, shift 1
154 can be quantified. The six parameters a ₁ ,. . . , A ₆ can be estimated by this process, where a ₁ and a ₄ correspond to the translation, and a ₃ and a ₅
DOO is corresponding to the rotation, and _{a 2,} and the _{a 6,} corresponding to the deviation. These parameters can be estimated for each sampling around the bounding box 1120. Average curl for each frame t
(curl) is, C (t) = _- is calculated as _a 3 + a _5.
This curl in each frame quantitatively gives the degree of rotation or spin of the finger skin about the pivot area. That is, the flow vector
An expression C (t) of the behavioral part of the obtained fingerprint, which is calculated from [u (x, y), v (x, y)] 1050, is obtained. Also, the average translation vector of the frame T = (a ₁ , a
₄ ) is calculated.

For all of the possible outcome biometrics described, we believe that conventional behavioral or physiological biometrics can be used. For the purpose of representation (template) and matching purpose of the resulting biometrics, these conventional biometrics are well understood in the prior art (see above for fingerprints by Lhasa, Chen and Jane, see above). thing). For the other part of the resulting biometrics, the behavioral part, we have left the user's motion as a function of some one-way or higher order time α (t). Matching of this part corresponds to matching of the stored template α ′ (t) with the function α (t). Such matching is again better understood in the prior art and is routinely performed in the field of signature identification. The following references provide examples of matching approaches. V. S. Narwa, "Automated On-Line Signature Identification", Proceedings of IEEE, pp. 215-239, 19
The above-cited references, February 1997, are fully included in the present invention.

Here, the obtained biometrics is
After being matched with the stored template, two “Yes / No” (214 in FIGS. 4 and 5) responses, or two scores S ₁ and S ₂ ( _{2 in} FIGS. 5 and 7)
31 and 233). Any of the methods for combining two biometrics described in FIGS. 4-7 may be used as conventional and user-defined biometrics of the resulting biometrics that can be used to make matching decisions. Can be used. The present invention is summarized below.

(1) an acquisition device for acquiring one or more sets of the biometrics from a subject over a time interval associated with a controlled change in each of the one or more biometrics; Storage means for storing the combined biometrics, which are the respective changes and the resulting biometrics, and the respective controlled changes with respect to time in a corresponding manner.

(2) The biometrics include physiological biometrics, behavioral biometrics, fingerprint, face, palm print, iris, retina, footprint, gait, signature, key
The system of (1), wherein the system comprises any one or more of a stroke pattern and a voice.

(3) The controlled change is:
The system according to (1), which is executed by the subject.

(4) The controlled change is:
Guided by an external mechanism to the subject,
The system according to (1).

(5) The system according to (4), wherein the mechanism includes at least one of a change in light, a change in light frequency, and a change in light intensity.

(6) The controlled change is induced by one or more of deformation, force, pressure, motion, torque, frequency change, gesture, energy change, volume, acceleration, and pattern of the biometrics. , (1).

(7) The method according to (1), wherein the biometrics is a fingerprint, and the controlled change includes one or more of finger movement, finger rotation, finger pressure, and finger force. The described system.

(8) The biometrics is a face, and the controlled change is facial movement.
The system according to (1).

(9) The biometrics is a face, and the controlled change is a face deformation.
The system according to (1).

(10) The system according to (1), wherein the biometrics is a palm, and the controlled change is palm movement.

(11) The biometrics is voice, and the controlled changes are volume, frequency,
The system according to (1), wherein the system includes any one or more of a pattern and intonation.

(12) The biometrics is a step, and the controlled change includes one or more of a stop pattern, a speed, a sway, a course, a posture, a hop, a skip, and a stride. (1) System.

(13) The system according to (1), wherein the biometric is a signature, and the controlled change includes any one or more of slope, loop, stretch, size, and interval.

(14) A method performed by a biometrics system, wherein the method comprises obtaining one or more of the biometrics from a subject over a time interval associated with a controlled change in each of the one or more biometrics. Obtaining a set of metrics; combining the biometrics and the controlled respective changes and the resulting biometrics, the combined biometrics and the combined biometrics and each controlled change. Associating and storing.

(15) an acquisition device for acquiring one or more sets of said biometrics from a subject over a time interval associated with a controlled change in each of said one or more biometrics; A computer system, comprising: a combined biometric, which is each controlled change and a resulting biometric; and storage means for storing the combined biometric and each controlled change in a corresponding manner. .

(16) obtaining one or more sets of said biometrics from a subject over a time interval associated with a controlled change in each of said one or more biometrics; A computer comprising a computer program for executing a combined biometric, each change and resulting biometric, and storing the combined biometric and each controlled change in a corresponding manner. A readable recording medium.

[Brief description of the drawings]

FIG. 1 is a diagram showing a conventional example of conventional biometrics.

FIG. 2 is a block diagram of an automated biometrics system for authentication.

FIG. 3 is a block diagram of an automated biometrics system for identification.

FIG. 4 shows the possibility of combining two biometrics at the system level,
The coupling is performed by D.

FIG. 5 shows the possibility of combining two biometrics at the system level,
The connection is performed by

FIG. 6 shows the possibility of combining two biometrics at the system level, AD
The connection is made by D.

FIG. 7 illustrates the possibility of combining two biometrics at the system level, wherein the combining is performed sequentially.

FIG. 8 is a concept for combining one biometric with a user's action (which is another biometric) to produce a biometric that is varied by a series of user-controlled changes. Block schematic diagram.

FIG. 9 is a view exemplifying fingerprint biometrics obtained by a user rotating a finger on a scanner according to a pattern;

FIG. 10 is a diagram illustrating fingerprint biometrics obtained when a user has four degrees of freedom to move a finger on a scanner.

FIG. 11 is a diagram illustrating an example obtained by generating a palm print image as a sequence.

FIG. 12 is a diagram illustrating an example obtained by generating a face image as a sequence.

FIG. 13 is a block diagram for extracting a behavioral part of the obtained biometrics, in which the extraction is performed from a fingerprint obtained by rotating as shown in FIG. 9;

FIG. 14 is a diagram showing a local flow computation block as a block base from a fingerprint image sequence obtained and input;

FIG. 15 is an explanatory diagram of calculating a bend or rotation of a finger, which is an active part of the obtained fingerprint, as a function of time.

FIG. 16 is an explanatory diagram of calculating a bend or a rotation of a finger, which is an active part of the obtained fingerprint, as a function of time.

[Explanation of symbols]

 110: signature 120: voiceprint 130: fingerprint 140: iris 200: automatic fingerprint authentication system 210: input 220: template 230: database

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Rudolph Marten Ball United States 10507, Nottingham Road 83, Bedford Hills, NY (72) Inventor Citra Doray United States 10541, Maple, New York Avenue 7 (72) Inventor Narini Kay Larsa, United States 20603, White Plains, NY, Granada Clesent 14, Apartment # 17

Claims (16)

[Claims] An acquisition device for acquiring one or more sets of said biometrics from a subject over a time interval associated with a controlled change in each of said one or more biometrics; and said biometrics and said controlled Storage means for storing the associated biometrics, which are each change and the resulting biometrics, and the respective controlled changes over time, in association with each other. 2. The biometrics comprises physiological biometrics, behavioral biometrics, fingerprint,
The system of claim 1, wherein the system includes any one or more of a face, palm print, iris, retina, footprint, gait, signature, keystroke pattern, and voice. 3. The system of claim 1, wherein the controlled change is performed by the subject. 4. The system of claim 1, wherein the controlled change is induced by an external mechanism to the subject. 5. The system of claim 4, wherein the mechanism comprises one or more of a light change, a light frequency change, and a light intensity change. 6. The controlled change is induced by any one or more of deformation, force, pressure, motion, torque, frequency change, gesture, energy change, volume, acceleration, and pattern of the biometrics. The system according to claim 1. 7. The biometrics is a fingerprint, and the controlled changes include any one or more of finger movement, finger rotation, finger pressure, and finger force.
The system according to claim 1. 8. The biometrics is a face,
The system of claim 1, wherein the controlled change is facial movement. 9. The biometrics is a face,
The system of claim 1, wherein the controlled change is a facial deformation. 10. The biometrics is a palm and the controlled change is palm movement.
The system according to claim 1. 11. The system of claim 1, wherein the biometrics is voice and the controlled changes include any one or more of volume, frequency, pattern, and intonation. 12. The method of claim 1, wherein the biometric is a gait and the controlled change includes any one or more of a stop pattern, speed, sway, course, posture, hop, skip, and stride length. The described system. 13. The system of claim 1, wherein the biometrics is a signature and the controlled changes include any one or more of slope, loop, stretch, size, and spacing. 14. A method performed by a biometrics system, the method comprising: receiving one or more biometrics from a subject over a time interval associated with a controlled change in each;
Obtaining a set of one or more of said biometrics; a combined biometric and said combined biometric, each control being said biometric and said controlled respective change and resulting biometric. And storing the corresponding changes. 15. An acquisition device for acquiring one or more sets of the biometrics from a subject over a time interval associated with a controlled change in each of the one or more biometrics; and the biometrics and the control. A computer system comprising: a combined biometric, which is each of the obtained changes and the resulting biometrics; and storage means for storing the combined biometrics and each controlled change in a corresponding manner. 16. Obtaining one or more sets of said biometrics from a subject over a time interval associated with a controlled change in each of said one or more biometrics; Computer-readable program comprising a computer program for performing a step of storing a combined biometric that is a change and a resulting biometric and the associated biometric and each controlled change in a corresponding manner. Recording medium.