KR101087833B1 - Fingerprint recognition method by learning effect - Google Patents

Abstract

PURPOSE: A fingerprint recognition method by learning effect is provided to prevent a malfunction by offering initial learning time and replacing the comparison fingerprint into a registered fingerprint. CONSTITUTION: A control/recognition part checks a registered fingerprint(S50). The control/recognition part compares and recognizes the comparison fingerprint which is recognized in the fingerprint image part with the registered fingerprint based on a feature point. The control/recognition part confirms whether the comparison fingerprint is generally admitted due to a learning effect during initial learning time(S53). If the confirmed comparison fingerprint group is maintained within an allowable error range, the comparison fingerprint is updated with a new registered fingerprint(S57).

Description

Fingerprint recognition method by learning effect

The present invention relates to a fingerprint recognition method using a variable method in consideration of the aging and learning effect, and if the status of the fingerprint at the time of measurement due to the user's habits and the fingerprint of one of the registered living body is different, it is repeated to learn a new registration fingerprint When the aging effect occurs, update the comparison fingerprint to the enrolled fingerprint to prevent malfunctions. It is about a recognition method.

In general, fingerprint recognition methods that are the mainstream of biometrics are classified into image-based fingerprint recognition and feature-based fingerprint recognition.

The image-based fingerprint recognition method is a method using ridge information. The overall directional information of a fingerprint image is applied by applying a technique such as a gaber filter, a fast Fourier transform (FFT), a tilt, a directional histogram, and a projection. One of the most classic ways to use.

The feature-based fingerprint recognition method is the most common method consisting of two processes, minutiae extraction and matching, and various image processing such as smoothing, separation of foreground and background areas, binarization, thinning, etc. Apply technique

The spatial features of the extracted feature points (minutiae) are used.

Meanwhile, the conventional fingerprint recognition method applying the above fingerprint recognition method to a personal device such as a smart card or a USB security token using low power stores only the fingerprint data on the smart card or the USB security token, and all fingerprint authentication operations are performed. The product at the level (Store-on-Card) is commercialized, and the method of performing the fingerprint authentication process on the smart card or the USB security token (Match-on-Card) is currently being studied and is illustrated in FIG. As is done.

Here, the fingerprint recognition method shown in FIG. 1 is a conventional fingerprint recognition method applied to a USB security token using low power, and is applied in the same manner to a smart card. In this case, the host computer below is replaced with a card reader.

Referring to FIG. 1, first, a host computer receives a fingerprint image of a user to be registered for user authentication, preprocesses the input fingerprint image (S10), and extracts feature points of the fingerprint from the input fingerprint image (S12). In addition, a feature data structure having geometrical information such as a relative angle difference between the extracted feature points, a relative distance, and the number of ridges is generated (S14), and the feature data structure is registered in the USB security token (S16).

As described above, when a predetermined fingerprint image (hereinafter, referred to as a comparative fingerprint image) is input to the host computer while the feature point data structure extracted from the fingerprint image of the user is registered in the USB security token, the host computer is configured to compare the fingerprint image. In the same manner as the user fingerprint registration process, preprocessing (S18), feature point extraction (S20), and data structure generation (S22) are performed to generate a feature point data structure for the comparative fingerprint image and transmit it to the USB security token.

Accordingly, the USB security token extracts the same feature point from the feature point data structure of the registered user and the feature point data structure of the comparative fingerprint image, such as parameters for rotation and transition between the extracted feature points, for example, two fingerprint images to be compared. After calculating the rotated angle (Δθ) between, the shifted coordinate amount (Δx, Δy) between the two fingerprint images to be compared (S24), after correcting the feature point for the comparison fingerprint image by the calculated parameters ( In operation S26, the user authentication is determined according to whether or not the registration fingerprint and the comparison fingerprint match by performing a point matching operation.

In this case, the USB security token performs a floating-point operation and a multiplication operation to correct and compare the feature points of the enrolled fingerprint and the comparison fingerprint to determine user authentication.

However, the conventional fingerprint recognition method applied to a personal device such as a smart card or a USB security token that uses low power as described above performs only a feature extraction process in a card reader or a host computer and a personal device such as a smart card or a USB security token. Because it is necessary to perform floating point and multiplication operations entirely in order to correct and compare the feature points of registered and comparison fingerprints, it is difficult to technically satisfy hardware resources, such as a large number of memories, that are required for complex operations. .

2 illustrates an embodiment in which a fingerprint recognition method is applied to a USB bus token. When the USB bus token is replaced with a smart card, the host is replaced with a card reader.

Referring to FIG. 2, the fingerprint recognition method extracts a feature point for user fingerprint information from a host and stores the feature point information in a personal device, and extracts a feature point for comparative fingerprint information from a host and saves the feature point for the fingerprint information. Comparing with the user fingerprint, the necessary parameters for calibration are calculated, and after the corrected comparison fingerprint, the authentication process consists of matching two fingerprints. The fingerprint image is performed by the fingerprint image unit 3 including a sensor for recognizing a fingerprint and a transducer for digitally converting the detected sensing value. The host is the name of the control recognition unit 5, the security token may be performed through the USB (USB), or may be configured to be included in the host.

The host extracts a feature point for a registration fingerprint in a registration process requiring a lot of floating-point and multiplication operations, or extracts a feature point for a comparison fingerprint in an authentication process and corrects the comparison fingerprint.

In the USB security token, only fixed-point and addition operations are repeated to calculate parameters for the feature point data structure of the registered fingerprint and perform a point matching operation for user authentication.

Fingerprint recognition method for a personal device using a low power according to the present invention made as described above is performed as follows.

Initially, the host receives and preprocesses a fingerprint image of a user to be registered for user authentication (S30).

After extracting the feature points of the registration fingerprint (S32), a feature point data structure having geometric information between the extracted feature points is generated (S34), and then the feature point data structure of the user is transmitted and registered by the USB security token (S36). .

In this case, when the host extracts the same feature point between the enrolled fingerprint and the compare fingerprint from the USB security token and calculates a predetermined parameter, the host receives the same fingerprint image twice to reduce the search area for the feature point. Only the feature points which are common among the feature points of the two fingerprint images obtained by repeating twice are extracted.

In addition, the feature point data structure may have geometric information with neighboring feature points in addition to the ridge end point and branch point in the fingerprint image, in order to extract the same feature point more efficiently. Referring to FIG. 3 and FIG. 4, the host has geometric information such as relative angle difference, relative distance, number of ridges, etc. between five neighboring feature points. From the registered fingerprint image, various image processing techniques such as image smoothing, background and foreground separation, binarization, and thinning are applied to extract feature points consisting of coordinates and direction information of the ridge end point and branch point.

For each reference feature point M, the Euclidean distance with respect to the coordinate values with the remaining feature points is obtained, and then, are sorted in ascending order, and the relative distance DMn with the other feature points is closest to M1. Five neighboring feature points Mn are determined from M5 to M5, and the relative angle difference θMn and the number of ridges RMn between each reference feature point M and the five neighboring feature points are calculated.

In this case, the relative angle difference θ Mn between the reference feature point M and the five neighboring feature points Mn is counterclockwise based on the direction of the reference feature point M, as shown in FIG. 4. Calculate

Accordingly, the distance between the five feature points closest to the general feature point data consisting of the coordinate values and direction information of the end point and the branch point as described above, the distance between the five neighboring feature points DMn, and the relative angle between the five neighbor feature points A feature point data structure with geometrical information of the difference (θMn) and the number of ridges (RMn) existing between five neighboring feature points is generated, and the table of feature point data structures thus created is shown in Table 1 below. can do.

M M n θ Mn D Mn R Mn 100.100.220.1 120.80.210.0 170 28.28 0 109.70.205.0 210 31.32 One 117.135.30.1 85 38.91 3 140.95.200.0 150 40.31 One 75.140.220.0 30 47.16 0

    As described above, when a predetermined comparison fingerprint image is input to the host while the feature point data structure of the user is registered in the USB security token, the host preprocesses the same as the user fingerprint registration process for the comparison fingerprint image (S38), A feature point data structure for the comparison fingerprint image is generated by performing a feature point extraction (S40) and a data structure generation (S42) procedure and transmitted to the USB security token (S42).

At this time, the process of extracting the feature points from the comparative fingerprint image and generating a predetermined feature data structure is performed in the same manner as the feature point extraction and the feature data structure generation process for the registered fingerprint image, as shown in Table 1 above. do.

Accordingly, the USB security token extracts the same feature point from the registered user's feature point data structure and the feature fingerprint data structure of the comparison fingerprint image, and the parameters for rotation and transition of the two images based on the same feature point, such as two to be compared The rotated angle Δθ between the fingerprint images and the shifted coordinate amounts Δx and Δy between the two fingerprint images to be compared are calculated, and the calculated parameters are encrypted and transmitted to the host (S44).

At this time, even if a parameter calculated by the rotation angle Δθ and the transition amount Δx, Δy transmitted to the host is leaked to the outside, only the parameter may be used to estimate the registered fingerprint registered in a personal device such as a USB security token. It becomes impossible.

In fact, in the USB security token, as shown in FIG. 5, the same feature point pairs M and M ′ are extracted from the feature point data structure of the registered fingerprint and the feature point data structure of the comparative fingerprint image, and then the pair of the same feature points M. A predetermined parameter is calculated by considering the rotation angle Δθ and the amount of transition Δx and Δy between and M ′ as the correction parameters for matching the coordinates of the two fingerprint images. In particular, the above parameter calculation can be performed by repetition of a simple addition operation without a multiplication operation, and thus can be operated in a personal device such as a USB security token or a smart card with limited hardware resources.

When the encrypted parameter is transmitted to the host as described above, the host corrects the feature point of the comparison fingerprint image by the encrypted parameter and transmits the same to the USB security token (S46). A point matching operation is performed on the feature points of the enrolled fingerprint to determine user authentication according to whether the enrolled fingerprint and the compare fingerprint match each other (S48).

In this case, the correction operation of the comparison fingerprint image is performed by matching the coordinate system of the comparison fingerprint with the coordinate system of the registered fingerprint by using the parameter value. The reason for performing the correction operation in the host is when performing correction on rotation. This is because iterative multiplication and floating point operations are required, and as a result fixed point floating point operations that occur when the correction operation on the comparative fingerprint image is performed on a personal device such as a smart card or a USB security token as in the prior art are performed. It can solve problems such as converting to arithmetic operations, performing many multiplication operations, generating errors, and increasing power consumption.

On the other hand, when the ridge of the fingerprint ends at the end point and the ridge of the fingerprint is referred to as a bifurcation point, the location (x, y-coordinate) and direction (θ) type (end point or branch point), and this data is stored in the memory to store the feature element along with the fingerprint image, and the stored enrolled fingerprint and the comparative fingerprint. Before the initial learning time t1 of the device, the fingerprint recognition quality of the enrolled fingerprint and the comparative fingerprint is lowered as shown in FIG. 6 due to the immaturity of the user, and the fingerprint recognition quality becomes constant when the initial learning time t1 elapses. It is called learning effect. The initial low fingerprint recognition quality is because the fingerprint recognition quality is different according to the proper position, pressing pressure, and finger humidity according to the recognition characteristics of the sensor. (t1) is required. Normal learning time may vary depending on the user's age or occupation, but usually takes 1 week to 2 months.

On the other hand, the aging effect as shown in Fig. 7 as a factor on the fingerprint recognition quality. This is because the recognition quality of the comparative fingerprint varies depending on the time, number, and time of the user's pressing. For example, the fingerprint recognition quality changes rapidly at the aging time (t2, t3, t4), which is called an aging effect. The aging effect changes towards younger fingerprints, and worse toward older. Therefore, the aging time t2, t3, t4 of FIG. 7 means a relatively long time compared to the learning time. However, the fingerprint reader currently produced has the inconvenience of judging that the fingerprint is different even by a true user by simply comparing the enrolled fingerprint and the comparative fingerprint without considering such learning and aging effects.

In addition, as a technology of the fingerprint recognition device, it is known as Korean Patent Registration No. 10-0914615, Domestic Patent Registration No. 10-0604324, Domestic Patent Registration No. 10-0499799, and Domestic Patent Registration No. 0-0347269 The technology itself is a reality that has a source of malfunction due to the positional change due to the user's fingerprint state or habit for fingerprint recognition.

The present invention is to solve this problem, an object of the present invention is to register a comparison fingerprint after the initial learning time to prevent the quality of the registered registration fingerprint and the comparison fingerprint at the time of the initial learning time to prevent malfunction of the learning immaturity of the device It is to provide a fingerprint recognition method by the learning effect to replace the fingerprint to prevent malfunction.

Another object of the present invention is to replace the comparative fingerprint with a registration fingerprint when the initial learning time due to the learning effect elapses, and to replace the comparison fingerprint with a registration fingerprint when the aging effect caused by environmental changes to increase the fingerprint recognition rate. It is to provide a fingerprint recognition method.

Another object of the present invention is to compare the registration fingerprint and the comparison fingerprint to determine whether the abnormality, differentiation of the registration fingerprint due to the difference in the dryness of the finger due to the change in the environment of the body, the season changes or the commute time zone and The aim of this study is to provide a fingerprint recognition method by learning effect that compares the corresponding comparative fingerprints and reduces recognition failure by considering the individuality of place, country or user.

To this end, the present invention recognizes a fingerprint in the fingerprint imaging unit, and compares the comparison fingerprint and the stored registration fingerprint recognized by the control recognition unit to recognize whether the same,

Confirming whether there is a registered fingerprint through the control recognition unit;

The comparison fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imaging unit are recognized by the control recognition unit based on the feature points, and the initial recognition is made after confirming whether the recognized comparison fingerprint is normal by the learning effect. Normal confirmation step of checking up to the learning time (t1);

A maintenance check step of checking whether the comparison fingerprint checked in the normal check step is kept constant within an error range;

If it is confirmed as maintenance, it provides a fingerprint recognition method based on the learning effect of sequentially performing the registration fingerprint updating step of updating the comparison fingerprint with a new registration fingerprint.

The present invention also recognizes the fingerprint in the fingerprint image unit, and compares the comparison fingerprint and the stored registration fingerprint recognized by the control recognition unit to recognize whether the same,

Confirming whether there is a registered fingerprint through the control recognition unit;

The comparison fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imaging unit are recognized by the control recognition unit based on the feature points, and the initial learning whether the recognized recognition fingerprint is normal by checking whether it is normal by the learning effect. A normal checking step of checking up to a time t1;

A maintenance check step of confirming whether the comparison fingerprint group checked in the normal check step is kept constant within an error range;

A registration fingerprint updating step of updating the comparison fingerprint with a new registration fingerprint if confirmed as maintenance;

Following the renewal phase, the change caused by the aging effect is checked to see if there is a change in the template (templete) by season or time zone. If there is no change, the comparison step is performed by checking the enrolled fingerprint and the comparison fingerprint. The present invention provides a fingerprint recognition method based on a learning effect, including a variation checking step of performing a second update step of replacing a comparative fingerprint with a new enrolled fingerprint.

As described above, the present invention has an initial learning time so that the quality of the registered enrolled fingerprint and the comparative fingerprint upon confirmation is prevented from malfunctioning due to immature learning of the device until the initial learning time, and after the initial learning time, the comparative fingerprint is registered as a registration fingerprint. Replace to prevent malfunction.

The present invention also replaces the comparative fingerprint with a registered fingerprint when the initial learning time due to the learning effect elapses, and increases the fingerprint recognition rate by replacing the comparative fingerprint with a registered fingerprint for each aging time when an aging effect occurs due to environmental changes.

The present invention recognizes a registered number of registered fingerprints and a set number of times when the comparative fingerprints at the time of confirmation are set to prevent errors and delays in determining whether an abnormality is caused by the aging effect of the comparative fingerprints due to environmental changes after the initial learning time has elapsed. When the aging effect occurs in the same pattern, the registration is updated with a new comparative fingerprint with an aging effect instead of the existing registration fingerprint, thereby reducing the time and reliability of fingerprint recognition.

The present invention differentiates the registration fingerprint due to the difference in the dryness of the finger due to the change in the environment of the body, the change of seasons or the time of commute when determining whether there is an abnormality by comparing the registration fingerprint and the comparison fingerprint, and based on this In order to reduce the recognition failure by comparing the corresponding comparison fingerprints, it should be accurate considering the individuality of the place, country or user.

1 is a view showing a conventional fingerprint recognition process,
2 is a view showing another existing fingerprint recognition process,
Figure 3 is an enlarged view showing an example of the reference feature of the fingerprint,
4 is a geometrical diagram between a reference feature point and a neighbor feature point of FIG. 3;
5 is a block diagram illustrating a process of extracting feature points of a comparative fingerprint based on a registered fingerprint;
6 is a graph showing a general learning effect,
7 is a graph showing a general aging effect,
8 is an exemplary system to which the present invention is applied;
9 is an exemplary enlarged fingerprint configuration illustrating the present invention;
10 is an exemplary field configuration included in a data header portion for fingerprint registration according to the present invention.
11 is a flowchart illustrating an example of the present invention;
12 is a flowchart illustrating another example of the present invention.

8 is a diagram illustrating a system configuration for carrying out the present invention. When a fingerprint is registered or is registered and is confirmed, the sensor is recognized by a sensor recognition unit, and a sensor unit such as a fingerprint recognition unit described in Korean Patent Registration No. 10-0604324, etc. 1) and a fingerprint imaging unit 3 having an analog digital converter 2 for converting a sensing value of the sensor unit 1 into a digital value;

Receiving a digital image value through the fingerprint imaging unit 3 to distinguish the neighboring feature point (Mn) in the relative dispersion distance between the reference feature point (M) and the reference feature point (M) as illustrated in Patent No. 10-0452324 A control recognition unit (5) for instructing to store the image judged and judged as a reference point and a comparison point in the memory unit 8 as a registration fingerprint or a comparison fingerprint;

A keyboard 4 for initial setting or command of the control recognition unit 5;

A display unit 6 for displaying the recognition and the instruction through the control recognition unit 5 on the screen;

It comprises the drive part 7 which drives a switch, a key, or the set drive means based on the recognition and command | indication by the control recognition part 5.

The fingerprint image unit 3 and the control recognition unit 5 may be installed in one device.

It may be installed separately by a separate wireless or wired method, or may be installed through an internet network.

9 is an enlarged fingerprint configuration example showing an example of a registration fingerprint 1 that recognizes a feature point for fingerprint recognition as a data value used in the present invention, a branch point JP1 having coordinate values X1 and Y1, and a coordinate value X2. End point EP2 having Y2, end point EP3 having coordinate values X3 and Y3, reference point RP4 having coordinate values X4 and Y4, direction of the point (θ1, θ1'θ2, θ3) , θ4) is illustrated.

FIG. 10 is an exemplary data diagram of registration fingerprint 1 in which each point JP1, EP2, EP3, RP4 shown in FIG. 9 is used as a position coordinate of a feature point, and is divided into fields including direction, type, relative distance, and quality. In this data field, the data of the sensor unit 1 is converted into a digital value by the analog digital converter 2 and then included in the header unit and sent to the control recognition unit 5, and the control recognition unit 5 sends the header of the corresponding data. The sections are classified and stored in the memory section 8 by the fields shown in FIG. In this case, the header part may be provided by the control recognition part 5.

FIG. 11 is a flowchart of the present invention, in which the fingerprint image unit 3 recognizes a fingerprint and compares the comparison fingerprint recognized by the control recognition unit 5 with the stored registration fingerprint to recognize whether the fingerprint is the same.

A confirmation step (S50) of checking whether there is a registered fingerprint through the control recognition unit 5;

The comparison fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imaging unit 3 are recognized by the control recognition unit 5 based on the feature points, and the recognized fingerprint is checked to determine whether it is normal by the learning effect. A normal confirmation step (S53) of confirming whether it is recognized as an initial learning time t1;

A maintenance confirmation step (S55) of confirming whether the comparison fingerprint group checked in the normal confirmation step (S53) is kept constant within an error range;

If it is confirmed as maintenance, the registration fingerprint updating step (S57) of updating the comparison fingerprint with a new registration fingerprint is sequentially performed.

If there is no registration fingerprint in the confirmation step (S50), performs a registration guide step (S51) to guide the new registration to register.

In the normal checking step (S53), if the comparison fingerprint is not normal by the learning effect, the comparison fingerprint (S54) is performed to compare the registration fingerprint and the comparison fingerprint to check whether there is an abnormality.

When the comparison fingerprint by the learning effect is not normal in the maintenance confirmation step (S55), the comparison confirmation step (S56) to first check whether the abnormality is compared by comparing the registration fingerprint and the comparison fingerprint.

12 is a flowchart illustrating another embodiment of the present invention.

The fingerprint image unit 3 recognizes the fingerprint, and compares the comparison fingerprint and the stored registration fingerprint recognized by the control recognition unit 5 to recognize whether the same,

A confirmation step (S50) of checking whether there is a registered fingerprint through the control recognition unit 5;

The comparison fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imaging unit 3 are recognized by the control recognition unit 5 based on the feature points, and the recognized fingerprint is checked to determine whether it is normal by the learning effect. Normal confirmation step (S53) to confirm up to the initial learning time (t1) to be recognized as normal;

A maintenance confirmation step (S55) of confirming whether the comparison fingerprint group checked in the normal confirmation step (S53) is kept constant within an error range;

A registration fingerprint updating step (S57) of updating the comparison fingerprint with a new registration fingerprint if it is confirmed as maintenance;

Following the update step (S57), check the fluctuation due to the aging effect of changing the comparative fingerprint (templete) by season or time zone, and if there is no change, perform the comparison step (S61) to check the registration fingerprint and the comparative fingerprint, and change If there is a change check step (S60) to perform a second update step (S63) to replace the comparison fingerprint with a new registration fingerprint.

If there is no registration fingerprint in the confirmation step (S50), performs a registration guide step (S51) to guide the new registration to register.

In the normal checking step (S53), if the comparison fingerprint is not normal by the learning effect, the comparison fingerprint (S54) is performed to compare the registration fingerprint and the comparison fingerprint to check whether there is an abnormality.

When the comparison fingerprint by the learning effect is not normal in the maintenance confirmation step (S55), the comparison confirmation step (S56) to first check whether the abnormality is compared by comparing the registration fingerprint and the comparison fingerprint.

Referring to the present invention performed in this way in conjunction with Figure 8,

For example, the fingerprint recognition unit 5 receives the fingerprint data as shown in FIG. 9 through the fingerprint imaging unit 3 and extracts feature points of the corresponding fingerprint (M in FIG. 4 and branch point JP1 in FIG. 9, for example). An example of this may be to extract the fingerprint data including the field as shown in FIG. 10. The registration fingerprint stored in the memory unit 8 is loaded in the field structure as shown in FIG. In step S50, the user performs a verification step (S50) of checking whether the comparison fingerprint having the feature point form as shown in FIG. 9 and the registration fingerprint provided by the fingerprint image unit 3. In this case, the registration fingerprint does not match. In this case, in the case of the fingerprint having a specific point as shown in Fig. 9, the storage of the corresponding fingerprint data in the field structure as shown in Fig. 10 will be described. Can be.

When there is a registration fingerprint in the checking step (S50), the control fingerprint unit 5 recognizes the comparison fingerprint having the feature points of the image recognized by the fingerprint image unit 3, and the control fingerprint unit 5 recognizes the registered fingerprint in the memory unit. In (8), it is checked whether or not there is a match, and at the same time, the position of the reference feature of the comparison fingerprint (M in FIG. 4, JP1 in FIG. 9, for example) is checked to see if there is a difference. To perform a comparison step (S54) to check, there is a difference in the position of the reference feature point (M) of the comparison fingerprint and the quality difference (for example, 1 in the quality of Figure 10, 2 grade 2, 3 grade 3, If it is 4, it is set as 4 grade and based on this, it is judged whether it is normal). If the enrolled fingerprint and the comparative fingerprint are within the same grade or tolerance range, the comparison fingerprint is compared to perform the comparison step (S54) to determine whether the same person as the person. The normal confirmation step (S53) is a process for determining whether the grade of the fingerprint measurement state is increased by the learning effect of the user, which is a feature of the present invention, and whether the normal quality is close to the first grade of FIGS. 6 and 7. To judge. Therefore, before reaching the first level, it is judged whether it is the same as the registered fingerprint in the state, and this is performed to perform the comparison step (S54) to determine whether it is the same to ensure the reliability of the device at an early stage.

Subsequently, the maintenance check step (S55) is performed to determine whether the fingerprint quality of one grade is confirmed after the normal check step (S53). This is to determine whether the quality persists considering that the learning effect is completed after reaching the first level of fingerprint status but there may be an error.

If the quality is not constant in the maintenance check step (S55), compare the registration fingerprint and the comparison fingerprint once in that state to determine whether the same or not (S56).

When the normality is maintained in the maintenance check step (S55), it is determined that the learning effect is completed, and the registration fingerprint update step (S57) is performed to update the comparison fingerprint with a new registration fingerprint, and waits for the next comparison fingerprint. Terminate the add operation.

The present invention may further perform the fingerprint verification process by the aging effect of performing additional steps (S60 to S63) as shown in FIG. 12 in addition to the fingerprint verification method by the learning effect.

That is, as shown in FIG. 12, the registration fingerprint updating step S57 is performed as it is. Subsequently, after the registration fingerprint updating step S57, a variation checking step S60 of checking a variation due to an aging effect shown in FIG. In the case of a young child, the state of the fingerprint improves as time goes by and the aging time (t2) is judged to be improved.In the case of an older person, the state of the fingerprint worsens as time passes. , t4).

If there is no change in the change checking step (S60) to perform a normal comparison step (S61) to check the registration fingerprint and the comparison fingerprint. However, if there is a change, it is determined based on the grade due to the aging effect, and when the set grade is reached, a second update step S63 is performed to prevent the malfunction by replacing the comparative fingerprint with a new registration fingerprint. For example, if a user's skin becomes dry, wet, aging or healthier according to the season, the comparison fingerprint by the aging effect is replaced with a new registration fingerprint to recognize that the quality of the registration fingerprint and the comparison fingerprint is different. This is to prevent malfunction by changing and to enable stable use.

Although the present invention has been described based on fingerprint recognition, it is also applicable to the iris recognition, which is another means of biometrics, by changing only the fingerprint image portion to the iris image portion.

1; sensor unit 2; analog digital converter 3; professional video unit 4; keyboard 5; control recognition unit 6; display unit 7; drive unit 8; memory unit

Claims (5)

The fingerprint image unit 3 recognizes the fingerprint, and compares the comparison fingerprint recognized by the control recognition unit 5 and the stored registration fingerprint to recognize whether the same,
A confirmation step (S50) of checking whether there is a registered fingerprint through the control recognition unit 5;
After comparing the fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imager 3 on the basis of the feature point in the control recognition unit 5, and after confirming whether the recognized comparison fingerprint is normal by the learning effect A normal confirmation step (S53) of checking whether it is recognized as an initial learning time (t1);
A maintenance confirmation step (S55) of confirming whether the comparison fingerprint group checked in the normal confirmation step (S53) is kept constant within an error range;
The fingerprint recognition method according to the learning effect, characterized in that to perform the registration fingerprint updating step (S57) to sequentially update the comparison fingerprint to a new registration fingerprint if confirmed as maintenance.
The fingerprint image unit 3 recognizes the fingerprint, and compares the comparison fingerprint and the stored registration fingerprint recognized by the control recognition unit 5 to recognize whether the same,
A confirmation step (S50) of checking whether there is a registered fingerprint through the control recognition unit 5;
The comparison fingerprint and the registration fingerprint of the image recognized several times by the fingerprint imaging unit 3 are recognized by the control recognition unit 5 based on the feature points, and whether the quality of the recognized comparison fingerprint is normal by the learning effect. A normal confirmation step (S53) of confirming whether it is recognized as normal by checking the initial learning time (t1);
A maintenance confirmation step (S55) of confirming whether the comparison fingerprint group checked in the normal confirmation step (S53) is kept constant within an error range;
A registration fingerprint updating step (S57) of updating the comparison fingerprint with a new registration fingerprint if it is confirmed as maintenance;
Following the update step (S57), check the change due to the aging effect to check whether there is a change in the template (templete) by season or time zone, and if there is no change, perform a comparison step (S61) to check the registration fingerprint and the comparison fingerprint. And a variation checking step (S60) to perform a second update step (S63) of replacing the comparative fingerprint with a new registration fingerprint if there is a change.
The fingerprint recognition method according to claim 1 or 2, wherein if there is no registration fingerprint in the confirmation step (S50), a registration guide step (S51) is performed to guide and register a new registration.
According to claim 1 or 2, if the fingerprint quality of the comparison fingerprint in the normal confirmation step (S53) is not the normal level due to the learning effect comparison step (S54) to check whether the abnormality by comparing the registration fingerprint and the comparison fingerprint Fingerprint recognition method by learning effect, characterized in that to perform.
The method according to claim 1 or 2, wherein in the maintenance check step (S55), when the comparison fingerprint by the learning effect is not normal, the comparison confirmation step (S56) is performed to first check whether the abnormality is compared by comparing the registration fingerprint and the comparison fingerprint. Fingerprint recognition method by learning effect, characterized in that to perform.

Images