Reference fingerprint data recording method, involves recording minimum of two fingerprint images that depict partially different areas of one finger of person

(54) NAME Device and method for registering reference fingerprint data (56) PUBLISHING QUOTES:

WO Al 0 184 494, WO Al 0 180 167 (57) SUMMARY. at_one fo_rf_{Aran (}j_e fs_r registration of reference fingerprint data for a person, at least two fingerprint images are recorded which depict at least partially different areas on one and the same finger of the person. Data from these images is stored in at least two alternative sets of reference fingerprint data to be used in checking the person's identity. This method makes it possible to use a small sensor that records only partial fingerprints, both when registering reference data and during verification, without the person having to verify, during the verification, that the sensor should register the same area of ​​the finger both when registering reference data and at verification.

Furthermore, a method of identity checking is described, wherein the method utilizes that there are alternative sets of reference fingerprint data, and devices for carrying out the methods.

<img file="SE526677C2_D0001.tif" />

The numbers in parentheses indicate the INID code.

526 677

<img file="SE526677C2_D0002.tif" />

SUMMARY

In a method for recording reference fingerprint data for a person, at least two fingerprint images are recorded which map at least partially different areas on one and the same finger of the person. Data from these images is stored in at least two alternative sets of reference fingerprint data to be used in checking the person's identity. This method makes it possible to use a small sensor that records only partial fingerprints, both when registering reference data and during verification, without the person having to verify, during the verification, that the sensor should register the same area of ​​the finger both when registering reference data and at verification.

Furthermore, a method of identity checking is described, wherein the method utilizes that there are alternative sets of reference fingerprint data, and devices for carrying out the methods.

2002-02-15 14: OS G: \ Fat \ CH \ aNS \ 2001 \ 2Q12717 SS! £&gt;) Submitted 1 2002-02-15.doc

526 677 • r

Field of the Invention

The present invention relates to a method and apparatus for recording reference fingerprint data for a person. It further relates to a method and apparatus for identity verification.

Background of the Invention

It has long been known that fingerprints can be used to check a person's identity. During the inspection, a current fingerprint is recorded from the person whose identity is to be checked and compared with previously registered reference fingerprint data. The check may refer to a verification of the person's identity. In that case, the current fingerprint is only compared with the reference data of the person whose identity is to be verified. The check may also refer to an identification of the person's identity. In this case, the current fingerprint is compared with the reference data for several different persons to determine if the current fingerprint comes from one of these persons.

There are different types of sensors for detecting fingerprints. The sensors may, for example, be optical, thermal, pressure sensitive or capacitive. Traditionally, the sensors have had a surface of a size that allows registration of complete fingerprints. However, in particular for capacitive sensors manufactured in silicon, the cost of a sensor is strongly dependent on the size of the sensor surface. Therefore, there is a desire to be able to reduce the sensors.

EP 0 813 164 discloses a thermal sensor which only allows registration of part of a fingerprint. To register a complete fingerprint, the user moves his finger over the sensor that records high frequency images. The images, which partially overlap,

526 677

<img file="SE526677C2_D0003.tif" />

are then pieced together to form a picture of a complete fingerprint. The identity check is performed in the usual way on the basis of complete fingerprints.

This sensor has the advantage of having a smaller sensor5 surface than sensors that record a complete fingerprint with a single image. The disadvantage, however, is that it is more difficult to handle for the user who has to drag his finger over the sensor at a relatively uniform speed for the image to become clear. If the user slowly pulls 10 or even stops the finger, the result deteriorates because the sensor is based on temperature differences that are smoothed out immediately if the finger is held still.

WO 00/49944, filed by the applicant for the present application, discloses the use of a small sensor that records a partial fingerprint, i.e., only part of a complete fingerprint. According to this application, a reference fingerprint image is registered by means of a traditional large sensor or by means of a small sensor, the images being pieced together into a picture of a complete fingerprint. In contrast, only an image of a partial fingerprint is registered in the identity check. This sub-image is compared with the considerably larger reference image in all conceivable positions to check whether the sub-image corresponds to any sub-area of ​​the reference image and thus can be assumed to come from the same person as the reference image.

This method allows the user to keep his finger on the sensor during the identity check. However, the comparison in all overlap positions requires a certain processor capacity and can take a relatively long time. Furthermore, essentially a complete reference fingerprint must be stored. This can be a disadvantage, for example, when the control is to be performed on a smart card which normally has both limited processor and memory capacity.

However, in other known methods of identity verification using fingerprints, complete fingerprints are normally not stored as reference fingerprint data

i. i ·: '&lt;

526 677 ·· ·· without the most relevant information is extracted from the fingerprint and saved as reference fingerprint data. For example, various sub-areas of an image of a fingerprint can be saved as reference data.

WO 01/84494, also filed by the applicant for the present application, further describes how identity control can be shared between a smart card and a processing unit to minimize the operations that need to be performed on the smart card and the required processor10 capacity. Specifically, the reference data in this case is divided into a public part, which is allowed to leave the smart card, and a private part, which is not allowed to leave the smart card. The public section contains a sub-area and coordinates indicating the location of additional sub-areas contained in the private section. An identity check records a current fingerprint for the person whose identity is to be checked. The public sub-area is sent from the smart card to the processing unit where it is compared to the current fingerprint to check if it matches anywhere in it. If so, the mutual positioning (translation and rotation) of the reference data and the current fingerprint have been determined. The coordinates in the public part of the reference data are then used to determine sub 25 areas of the current fingerprint to be matched to the sub areas of the private part of the reference data. These sub-areas of the current fingerprint are sent to the smart card where the identity check is terminated. As an alternative to the public part area, a set of special features or features, so-called features, can be stored in the public part of the reference data and used to determine the mutual location of the reference data and the current fingerprint.

If a small sensor, ie a sensor that detects only a partial fingerprint, is used for this application, the problem arises that the user may accidentally put his finger on the sensor in such a position that only

526 677

<img file="SE526677C2_D0004.tif" />

some or none of the sub-areas are within the sensor surface. If this happens, identity checks will not succeed because there are not enough sub-areas to compare. The user is then allowed to retry the finger in a new position. A number of attempts may be required and this year is of course troublesome for the user.

SE 0102376-1, which was filed on 29 June 2001 by the same applicant as in the present application and which was therefore not public at the time of this application, proposes a solution to this problem which consists in checking how well the reference data are aligned with the current application. fingerprint. Specifically, on the basis of where the public sub-area is located in the current fingerprint, it is possible to calculate, by means of the coordinates of the private sub-areas, whether the private sub-areas are within the sensory area and if the identity check thus has the prerequisites to succeed. If not, an indication can be given to the user on how to move his finger relative to the sensor to improve alignment.

Summary of the Invention

It is an object of the present invention to solve the problem that a user may need to make several attempts to identify himself with his finger in various positions on the sensor when it is of the type that records only a partial fingerprint.

The object is achieved in whole or in part by means of a method and a device for recording the reference fingerprint data according to claims 1 and 14 and a method and a device for identity checking according to claims 16 and 24, respectively.

Thus, the invention relates to a method for recording reference fingerprint data for a person, comprising the steps of receiving at least two fingerprint images that map at least partially different areas on one and the same finger of the person, and storing data

526 677

<img file="SE526677C2_D0005.tif" />

from said at least two fingerprint images as at least two alternative sets of reference fingerprint data, which includes data from at least partially different areas of the finger, for use in checking the person's identity.

As more than one set of reference fingerprint data is stored, a larger portion of the person's fingerprints are stored, thus increasing the chances that a partial fingerprint recorded during identity verification is within the mapped part of the fingerprint, which in turn leads to fewer, and optimally only one, attempts need to be made to, for example, verify a person's identity.

Instead of the user having to move around his finger and try in several different positions on the sensor at each time to verify his identity, this now only needs to be done once and that is when registering reference data. This should be a significant improvement for the user.

In addition, by storing alternative sets of reference fingerprint data, the same algorithms can be used both for a small sensor that detects a partial fingerprint and for a large sensor that records a complete fingerprint and which only needs a set of reference fingerprint data because the problem of finger position on the sensor is less.

The alternative sets of reference fingerprint data can be stored in different ways. For example, they can be stored as completely separate sets that do not share any data at all. Thus, if overlapping data is included in the alternative sets, this data is stored once for each set. The different sets may alternatively constitute different subsets of a jointly stored amount of data. Thus, if overlapping data is included in the alternative sets, it needs to be stored only once in this case.

526 677 ··· ··· · · ·· ···· ·· ····

The different sets of reference fingerprint data can be used in an identity check. Specifically, according to the invention, a method of identity checking comprises the steps of receiving a current fingerprint image which depicts a portion of a finger of a person whose identity is to be checked; comparing data from the fingerprint image with data from at least one first set of reference fingerprint data of at least two previously stored sets of reference fingerprint data relating to at least partially different regions of one and the same finger of a person whose identity is known; and, on the basis of the result of the comparison, determine which of said at least two alternative sets of reference fingerprint data to be used for the identity check. In the identity check it is therefore decided whether the person whose identity is to be checked has the same fingerprint as the person whose identity is known, ie if they are one and the same person.

Each of the sets of reference fingerprint data may comprise a plurality of subareas of the corresponding fingerprint image. This enables storage and matching of the reference fingerprint data on, for example, smart card and other devices with limited processor and memory capacity.

Furthermore, one of the sub-areas can be used to determine which set of reference fingerprint data is to be used in an identity check. Thus, no exhaustive control over all data in all sets of reference data need be made and thus the time for identity verification is extended very little even though there are several different sets of reference fingerprint data. This can be a special advantage when the identity check is done in contexts where, from the outset, there is a risk of slowing down, such as identity check on smart card, via networks or in databases where the check is to be made against reference data from several people.

526 677

<img file="SE526677C2_D0006.tif" />

As an alternative to using a sub-area to determine which set of reference fingerprint data to use in an identity check, features can be used. As mentioned above, a feature is a particular feature or feature of the fingerprint, for example, a point where a line ends or a point where a line divides. Such features are also known by the term minutia points. In this case, the positions and / or other information, such as the angle or type of feature, is extracted from the fingerprint. This information is then stored in the public part of the reference fingerprint data instead of the public part area.

The methods of the invention can be implemented in devices comprising a signal processor. The signal processor may be a processor provided with appropriate software for performing the procedures. It may also include application-specific hardware, such as an ASIC, a programmable logic circuit, such as an FPGA, or digital or analog circuits or any suitable combination thereof.

Brief description of the figure

The present invention will now be described in more detail by way of example with reference to the accompanying drawings, in which Fig. 1a shows schematically a fingerprint in which different areas mapped when registering alternative sets of reference fingerprint data are highlighted.

Fig. 1b schematically shows the same fingerprint as in Fig. 1a but with only one depicted area marked.

Fig. 1c schematically shows the same fingerprint as in Fig. 1a, but in which an area depicted during identity check is marked.

Fig. 2 schematically shows a system in which the methods and devices according to the invention can be used.

526 677

<img file="SE526677C2_D0007.tif" />

Fig. 3 is a flow chart showing an example of a method for recording reference fingerprint data.

Fig. 4 is a flowchart showing an example of an identity check method.

Detailed description of an embodiment

The present invention is based on the idea of ​​recording at least two different sets of reference fingerprint data from one and the same finger, corresponding to at least partially different areas of the finger. This idea is illustrated in Fig. 1a, which schematically shows a complete fingerprint 1, which is recorded from the outermost finger joint of a person's finger and in which is marked three areas 2a-2c which are imaged with a small sensor when recording the reference fingerprint data. The three areas correspond to three partial fingerprints. As shown in Fig. 1a, the regions 2a-c can overlap.

However, they do not have to.

As has been shown at the outset, the entire regions 2a-2c are not normally stored as reference fingerprint data but only extracted data. In Fig. 1b, the same fingerprints are shown as in Fig. 1a, but only the upper region 2a is marked. In this area, six different sub-areas AF are indicated. These subareas can be stored as reference 25 fingerprint data.

When the identity check is to be performed, it is usually sufficient that the user places his finger once on the small sensor. Fig. 1c again shows the same fingerprint 1 as in Fig. 1a. In the fingerprint, a current area 3 is indicated by the small sensor.

This current range is best matched with the topmost region 2a of FIG. 1a, which can be determined, for example, by comparing a sub-region of each of the regions 2a-c in FIG. 1a with the current region.

This in turn means that it is a set of reference fingerprint data that essentially corresponds to the

526 677

<img file="SE526677C2_D0008.tif" />

the top reference area to be used for the actual identity check.

If only the middle or the lower region had been mapped when registering reference fingerprint data, the identity check would not have succeeded as the current area would not overlap the area mapped when registering reference fingerprint data and it would therefore not be possible to determine if it came from the same person.

The user had to try at least once more with his finger in another position on the sensor.

Figure 2 shows a system that can, but does not need, be used for both reference fingerprint data registration and identity verification. The system comprises a fingerprint sensor 21, a computer unit 22, which is an example of a signal processor, a smart card reader 23 and a smart card 24.

The sensor 21 may be optical, capacitive, thermal, pressure sensitive or of any other suitable type. Its sensor surface is such that it can only detect a partial fingerprint from the outermost joint of a normal person's finger. A sensor for recording a complete fingerprint has at least a size of 200 * 200 pixels with 500 dpi resolution. Thus, with such a sensor, an image of the entire area of ​​the outermost finger joint that comes into contact with a substrate is recorded when the outermost finger joint is pressed against a support having equal or greater spread than the finger joint. A small sensor for a partial fingerprint may have a size of, for example, 128 * 128 pixels or less. This sensor will only image a portion of the area of ​​an outer finger joint that comes into contact with a support when the finger joint is pressed against it. Sensor 21 may be standalone or integrated with computer unit 22 or smart card reader 23 or smart card 24.

526 677 ··

The computer unit 22 may be an ordinary general computer, for example a PC. Alternatively, it may be a computer unit 22 which is dedicated for processing fingerprints. It may, for example, include a processor with program memory and working memory or a custom hardware such as an Application Specific Integrated Circuit (ASIC) or a programmable logic circuit such as a Field Programmable Gate Array (FPGA) or digital or analog circuits or any suitable combination thereof.

The smartcard reader 23 can be any commercially available smart card reader. Alternatively, it can be a customized smart card reader. It may be a standalone unit or integrated with the computer unit 22.

The smart card 24 can be any type of smart card on which reference fingerprint data can be stored. The smart card has a signal processing unit comprising a processor 25, a memory 26 for storing reference fingerprint data extracted from a reference fingerprint from the smart card holder, and a working memory 27, and communication circuits 28 enabling communication between the smart card reader 23 and the smart card 24. The communication circuits 28 may, but need not, require contact between the smart card and the reader.

The system may further comprise indicating means 29 which are intended to convey information to the user. The display means 29 may, for example, consist of a display on which messages or signals are displayed, by LEDs or some form of sound generation circuit. In Fig. 2, the display means are shown on the computer unit 22, but they can also be arranged on the sensor 21.

The following describes how registration of the reference fingerprint data can be carried out by means of the system in Fig. 2 and with reference to the flow chart in Fig. 3. In this example, the data extracted from the images of sub-areas and the reference fingerprint data are stored on a smart card.

···

526 677

<img file="SE526677C2_D0009.tif" />

follow one

The reference fingerprint data is otherwise referred to as a template. A template may include complete or partial fingerprinting in the raw or untreated form in which it is recorded. Normally, however, a template contains selected, processed and compressed data from the registered fingerprint. The latter year the case in this example.

The method begins with the user, i.e., the person for whom reference fingerprint data is to be registered, placing the outermost point of his finger on sensor 21. An image in gray scale of a region, e.g., area 2a of Figure 1, of the fingerprint is recorded, step 30. The image is a digital representation of the fingerprint area.

In the next step 31 it is checked whether the image is acceptable. This check may include, for example, one or more of the following steps.

For example, the computer unit 22 can check the recorded image to ensure that there is really a fingerprint in the image, that the fingerprint occupies a sufficiently large portion of the image and that the fingerprint is sufficiently clear.

The computer unit 22 can further check whether the user has applied the finger with sufficient pressure to the sensor and that any moisture on the user's finger has not made it impossible for the sensor 21 to distinguish between ridges and valleys in the fingerprint.

In addition, the computer unit 22 can binarize the recorded image and compare it with one or more previously previously recorded images to check if the last recorded image differs sufficiently from the previously registered images. For example, if one is to register three sets of reference fingerprint data on the basis of three reference images, these should preferably differ so much that they together cover a much larger area than a single image, so that the possibilities thereby increase that a later, in connection with

2772 77-7- ii: 7 77; : lli · 7 '¹'·· AK,&gt;' 7 t CCl \ 7:, „7, · '' 7 '2 ·' 'Pj 2772 j2 .hli

526 677 '·· * image occupied with identity check, such as the image of area 3 of Fig. 1c, is within the range covered by the reference images so that the identity check has the prerequisites to succeed.

If the registered image cannot be accepted, the computer unit 22, via the indicating means 29, sends a signal to the user about this, step 32. In the simplest case, the signal can only be a signal that the image has not been accepted. However, the signal may also contain an indication of the reason why the image cannot be accepted and / or an indication of what the user should do to remedy the problem. For example, if the unaccepted image overlaps with previously recorded images, the user may get an indication of the direction in which the finger is to be moved. For example, the computer unit 22 can calculate this by comparing a sub region of a previously registered and accepted image with the last registered image and determining where in the last registered image the sub region fits. In this way, the mutual position between the last recorded image and the previously registered and accepted image is determined and thereby the computer unit 22 can determine how the finger is to be moved relative to the sensor 21 in order to reduce the overlap.

Thereafter, the flow returns to step 30 for registration of a new image.

If the registered image is acceptable, in the next step, 33 sub-areas are to be included in the template.

First, a sub-area, hereinafter referred to as a public sub-area, is selected by the image for storage in a public part of the template. The sub-area can be selected in different ways. One way is to use the following three quality criteria: 1) Distinction, ie how easy a sub-area is to binaryize,

2) Uniqueness, ie how unique a sub-area is, and

3) geographical location, ie where a sub-area is located in the fingerprint.

526 677 ··

For example, the uniqueness can be controlled by correlating the subarea with the environment and choosing a subarea with little correlation with the surrounding. Alternatively, one can find sub-areas with one or more features.

With regard to the geographical location, sub-areas in the center of the image are preferred for where the fingerprint becomes least deformed when the user presses the finger with different pressures towards the sensor and furthermore, there is a greater chance that the sub-areas are included in a later registered image to be used for identity verification.

Depending on the relationship between the size of the sensor, the number of sub-areas and the size of the sub-areas, the aforementioned criteria may become more or less significant. The sub-areas may overlap each other and even

covers essentially the entire image. However, it may be advantageous to select sub-areas as the same algorithms for recording the reference fingerprint data and for identity verification can then be used for images from different large sensors.

As an alternative to selecting sub-areas with different quality criteria, fixed sub-areas can be selected.

In any case, a sub-area is selected to form the public sub-area. Preferably, only one public part area is selected so that as little information as possible about the user's fingerprints is available in the public part of the template. However, more public sub-areas can be selected to provide a more secure correlation of the public portion of the template to a current fingerprint and thereby obtain a more accurate orientation of the template relative to the current fingerprint.

When the public sub-area is selected, at least one but preferably several sub-areas, hereinafter referred to as "private sub-areas," are determined or stored for storage in a private part of the template on the smart card 24.

The private sub-areas are preferably selected according to the same quality criteria as the public sub-area (s).

526 677

<img file="SE526677C2_D0010.tif" />

Preferably, four to six private sub-areas are selected. More or fewer sub-ranges can be selected depending on the desired level of security, the desired speed of matching on the smart card 24, and available processor capacity on the smart card 24.

The size of the selected public and private sub-areas in this example (500 dpi) is 48 x 48 pixels, but can be easily adapted by one of ordinary skill in the art to meet the needs.

When the private sub-areas are selected, their location is also determined in relation to a reference point. For example, the reference point can be chosen to be the center point in the public sub-area or in one of these if they are multiple. The point of reference of the public sub-area thus defines the origin of the local coordinate system in which the positions of the private sub-areas are expressed. Other well-defined reference points, of course, can also be selected, for example with the help of features. The locations of the private sub-areas are specified as coordinates, for example the midpoint of the private sub-areas, relative to the reference point. These coordinates are stored as part of the public part of the template.

In step 34, the computer unit 22 then stores the template pending transfer to the smart card 24. The template's public portion (s) will thus contain the public sub-area (s) and coordinates of the private sub-areas relative to a reference point. Its private part will contain the private part threads. In the private section, comparison criteria can also be stored in the form of threshold values ​​for the degree of conformity to be achieved when matching the private sub-areas to sub-areas of the current fingerprint in order for the template and the current fingerprint to be considered to come from the same individual. The sub-areas are preferably stored in the form of bitmaps.

526 677 *·

Then, in step 35, it is checked whether the registration of the reference fingerprint data is complete. For example, the criterion may be that a predetermined number of alternative templates have been stored and / or that templates corresponding to a fingerprint area of ​​a predetermined size have been stored.

If the computer unit 22 finds that the registration is not complete, a signal is then given to the user, step 36. The signal may include, for example, an indication that the registration is not complete, an indication of how much of the predetermined area has been stored, or an indication of how the user must move his finger to proceed with the registration. The flow then returns to step 30 and the computer unit 22 waits for a new image to be recorded.

If the computer unit 22 instead finds that the registration is complete, all registered templates are transferred and stored in the memory of the smart card, step 37. They will then constitute alternative templates for one and the same finger and one person for selective use in subsequent identity checks.

Then a signal is given to the user that the registration is complete, step 38.

Reference fingerprint data has now been recorded for the user. Since the registration is done under safe conditions and the user's identity is also registered on the smart card, his identity is known.

In the above described embodiment, each recorded image is processed separately and a template or set of reference fingerprint data for each image is determined. The images thus do not need to overlap and the coordinates of the private sub-areas are expressed in a local coordinate system for each template where the public sub-area defines the origin. A variant of this is to gradually build up the alternative templates. In such an embodiment, the first image may be recorded and processed in the manner described above. When treating ···

526 677 ·· ·· ·· * · · • 5 ·* ·· ····

<img file="SE526677C2_D0011.tif" />

• · • ·· The completion of this image is clear, thus, a first template is stored containing a public sub-area and a number of private sub-areas. Once the second image has been registered and accepted in step 31, one or more of the sub-areas of the first template is compared with the second image to check if any of the sub-areas matches in the second image so that the position of the first and second images can thereby determined. If none of the sub-areas of the first template matches in the second image, the user will be given an indication that the finger will be moved and a new image will be registered. If any of the subareas matches, subareas are extracted from the second image in the same way as for the first image. The position of these subareas relative to the reference point in the first image is now known since the position of the first and second images has been determined. The positions can thus be expressed in a coordinate system common to the first and second images and where, for example, a sub-area extracted from the first image defines the origin. The sub-areas of the second image are thus stored with their coordinates in the first template, after which a check is made as to whether the registration is complete. If not, a third image is recorded and the procedure for the second image is repeated. When the registration is deemed to be completed, a single template is stored with a large number of sub-areas, of which only one is designated as a public sub-area at that stage. However, before storing the smart card in step 37, one or more additional sub-areas are designated as public sub-areas. Subsequently, the sub-areas can be divided and stored on the smart card as a number of separate templates, each containing a public sub-area and a number of private sub-areas with associated coordinates. Each template then contains a complete set of data to be used for verification. However, some data may be identical for two templates, for example, if two templates have a private one

526 677

<img file="SE526677C2_D0012.tif" />

<img file="SE526677C2_D0013.tif" />

sub-area which is the same for both templates. Alternatively, all the subsets can be stored together as a super template, where different subsets form the alternate templates. Each such subset consists of a public subdivision with a plurality of associated private subdivisions. Some private sub-areas may then belong to multiple subsets or templates, but are only stored once. In this case, the sub-areas can be stored in a table showing which sub-areas are public and which are private, and the respective sub-areas' coordinates in the common coordinate system where one of the public sub-areas defines the origin.

In yet another embodiment, the images are not processed sequentially, but first a number of registered images or a selection of these are pieced together into a large image. The puzzle can be done in some known way, for example by scoring on the basis of similarity between overlapping pixels, or by some correlation technique. From this linked image, subareas are extracted in the same way as described for the individual images above. One of the sub-areas is designated as a public sub-area with a reference point, in relation to which the positions of the other extracted sub-areas can be determined. When all sub-areas are extracted, additional sub-areas can be designated as public sub-areas. Both here and in the embodiment described above, if possible, the additional public sub-areas are selected with such density that, regardless of the position in which an image is taken of the outermost finger joint, there should be one, but preferably not more, public sub-areas that match in the image. The storage is then done as separate templates or as a super template.

The advantage of storing the templates as subsets of a super template is that for a given memory space, a much larger number of alternative templates can be stored since each sub-area is stored only once.

The following describes how an identity check which is a verification can be performed with the system i

526 677 ·· » * ··

I · »·

<img file="SE526677C2_D0014.tif" />

2, with reference to the flowchart of FIG. 4 and on the basis of the reference fingerprint data recorded by the method of FIG. 3, which in this case thus includes sub-areas stored on the smart card 24.

When a user wants to verify his identity, he or she places his smartcard 24 with the memory stored templates in smart card reader 23 and places the same finger used in registering the template on sensor 21.

When the user places his finger on the sensor, a current image, step 40, of an area on the outermost finger joint is recorded. The area may, for example, be the area 3 in Fig. 1c. The image is a digital representation of this area.

In a subsequent step 41, it is checked whether the current image is acceptable. The control may comprise one or more of the steps mentioned above in connection with step 31.

If the current image cannot be accepted, a signal is then given to the user in step 42 in the same manner as in step 32. Thereafter, the flow returns to step 40 and the computer unit 22 awaits the registration of a new image.

On the other hand, if the current image is acceptable, the computer unit 22 reads the public parts of the templates stored on the smart card 23, step 43. Each of the public parts includes, as has been stated above, a public part area and coordinates for the location of the private part areas.

In step 44, each of the public sub-areas is compared or correlated with the registered current image of the fingerprint area. The correlation can be made to the whole image or to a portion of it of predetermined size, preferably in the middle of the image. In the correlation, the public sub-area is swept over the image and in each position a pixel-by-pixel comparison is performed. About a pixel in «··

526 677

<img file="SE526677C2_D0015.tif" />

the template portion area corresponds to a pixel in the image of the current fingerprint, so a certain value is added, for example 1 to a sum. If the pixels do not match, the sum is not increased. When the template's public area has swept over the entire image or selected area, a position is obtained where the template's public area best correlates with or overlaps the current fingerprint. The public sub-area can also be rotated relative to the image of the current fingerprint to determine if a better correlation can be obtained.

On the basis of the public sub-area for which the highest score is obtained, in step 45, the template to be used in the final verification on the smart card is selected or determined. However, the score for the public subarea must exceed a predetermined reference sum, step 46. If not, the verification is considered unsuccessful and signal is thereby given to the user in step 47 in the corresponding manner described above, after which the flow returns to step 40.

The correlation can fail for at least two reasons.

Either the current fingerprint does not originate from the person or the finger from which the reference fingerprint was registered and then there simply is no equivalent to any of the public sub-areas of the current image. Alternatively, the actual fingerprint and template originate from the same person and the same finger, but the person in question holds his finger in such a position relative to the sensor 21 that none of the areas corresponding to the public sub-areas is within the sensor surface. However, since several alternative templates are stored, the probability of this is small.

On the other hand, if the templates are stored as separate templates and not as subsets of a super template, the current area that has been registered does not fully comply with any of the templates. «· • in ΓΛ ·· ···· adjacent templates.

but without, for example, two overlaps Some of the private sub-areas of the selected template will then have no equivalent in the current image. That being the case, the computer unit can determine with the help of the coordinates in the public part of the selected template, which coordinates thus indicate the location of the private sub-areas relative to the public sub-area. If such a case is detected, the computer unit may signal to the user that he or she should fine-tune the position of the finger on the sensor, step 48, which is indicated by dashed lines to indicate that this step is not always performed. Preferably, then, this signal includes an indication in which direction the fine tuning is to be made. When the fine-tuned image is recorded with the sensor 21 and the computer unit 22 checked so that a required number of the private sub-areas have equivalents within the current image, sub-areas of the current image are sent to the smart card 24 for final verification, step 49. Specifically, it is selected in the In the present image, a sub-region of a predetermined size around each point defined by the coordinates of the public part of the template. However, the subareas in the current image may be slightly larger than the corresponding subareas in the template to compensate for any deformation of the fingerprint if the finger is placed with different pressure on the sensor when recording the current image. Along with the sub-areas of the current image, an indication, such as a number or a coordinate pair, is also transmitted, as needed, for which of the templates on the smart card 24 to be used in the final verification.

If the templates are stored as separate templates, the public portion of the respective template contains the coordinates of all the private subsets of that template.

If the templates are instead stored as subsets of a super template, each public sub-area can be

526 677 ’ * *** ’··’^,}·· *^!·.· .:1 ’ :**:

private sub-areas in associated coordinates for all super templates or for a number of them. It may then happen that the coordinates of some of the private sub-areas are such that the corresponding area of ​​the controlled person's finger is not within the surface of the sensor and thus is not included in the current image that has been recorded. There is then no sub-area in the current image to send to the smart card. You can then send random data / noise instead of subareas to indicate that the subarea is missing or you can choose to send only the subareas that are within the current image and instead indicate what those areas are, for example using the coordinates. In this case, therefore, which template, i.e., which of the sub-sets of the super 15 template, is to be used by the data returned to the smart card 24 from the computer unit 22. In this context, it should be noted that the area of ​​the current image that matches the public sub-area can but does not need to be sent to the smart card.

On the smart card 24, then the final comparison is made between the private sub-areas of the selected template and the current sub-areas of the current image that have been sent to the smart card. The comparison does not form part of the present invention and is therefore not described herein.

Instead, reference is made to the above-mentioned WO 01/84494.

Alternative embodiments

Although a particular embodiment has been described above, it will be apparent to those skilled in the art that many alternatives, modifications and variations are possible in the light of the foregoing description.

For example, the smart card 24 and the smart card reader 23 may be another unit with which the computer unit 22 communicates. This second unit, which may be portable or stationary, need not be physically present by the computer unit 22, but the communication can take place, for example, via computer networks, telecommunications networks or wiring. For example, the other device may be another computer, one

526 677

<img file="SE526677C2_D0016.tif" />

<img file="SE526677C2_D0017.tif" />

·· ·· »* ·· mobile phone or PDA. The template can be stored in a protected memory, ie a memory protected against unauthorized access, in the other device.

In the description above, it has further been stated that when the correct template is to be determined during the identity check, all the templates' public areas are compared against the actual partial fingerprint. This is not absolutely necessary. It is conceivable to compare one public sub-area at a time against the current partial fingerprint and to check, after each comparison, whether the public sub-area matches well enough. In this case, the comparison can be interrupted and a template containing the public sub-area can be used for identity verification. Another alternative is to include in the public part of each template information about the relative location of the other theme sub-areas. Once a public sub-area that matches the current partial fingerprint has been found, then the template whose public sub-area is best placed relative to the current partial fingerprint can be determined directly. This is then the template that will be used for the verification itself.

In the example above, three templates were stored. It is conceivable to use more or less.

The identity check procedure can be used for both verification and identification.

In the description above, it has been stated that the public part of the template contains a sub-area of ​​a registered fingerprint image. Alternatively, however, the public part of the template may contain a set of features, for example, features that are retrieved from the same area of ​​the finger as the private sub-areas of the template. These can be used in a similar way to the public sub-area. Thus, when deciding which template to use for identity verification, the set of features in the public part of the template is sent to the computer unit, in which the set

526 677 ···

I * ··

<img file="SE526677C2_D0018.tif" />

···« ·· ··

<img file="SE526677C2_D0019.tif" />

is compared with the current fingerprint image. The template for which most matching features are obtained is then used for identity verification. The coordinates in the public part of the template, which may, for example, refer to a reference point in the form of a specific feature, are then used to determine which sub-areas of the current fingerprint image to be sent to the smart card or other unit where the identity check is to be performed.

As a further variant, features from the entire finger area covered by all templates can be registered and form public data common to all templates. When the identity check is sent, this public data is sent out from the smart card to the computer unit and compared with features in the current image. Only some features in the public data will have similarities in the current image, but with these, the current image can be aligned with the data on the smart card so that the appropriate template can be used for identity verification.

Figure 1a shows that the areas overlap.

This is not necessary, especially if the user receives a signal indicating how to move the finger after a failed verification attempt.

In the process described in connection with the flow chart in Fig. 3, it is stated that the reference fingerprint data is recorded successively until a certain criterion is met. As an alternative, one could allow the user to move around the finger to a plurality of different positions on the fingerprint sensor, register an image for each position, and when a certain number of images are recorded or after a certain period of time stop recording the images, selecting a predetermined number of images based on the overlap so as to achieve the best possible spread and determining and storing a set of reference fingerprint data for each of the selected images.

526 677 ···· ····