EP1604325A2

EP1604325A2 - Method for identifying persons and system for carrying out said method

Info

Publication number: EP1604325A2
Application number: EP04712589A
Authority: EP
Inventors: Florence Guillemot; Bernard Didier
Original assignee: Sagem SA
Current assignee: Idemia Identity and Security France SAS
Priority date: 2003-02-20
Filing date: 2004-02-19
Publication date: 2005-12-14
Also published as: WO2004077200A3; FR2851673A1; WO2004077200A2; US20070067639A1; FR2851673B1; US7583822B2

Abstract

According to the inventive method, and with the aid of the system, a sensor (20) is used to capture biometric characteristics of a person to be identified; identification data is extracted and the identification data of the person to be identified is compared to reference data in a data base (40), capturing an image of the folds of the skin (10) of the person to be identified, wherein the reference skin fold data of is acquired and stored in the base prior thereto; a first recognition is carried out using the image in a filtering step (103), prior to filtering (104) the reference data of the data base relating to other biometric characteristics, and the recognition (105) is continued with the aid of said characteristics. .

Description

Method for identifying people and system for implementing the method.

Today, methods of identifying or authenticating people are implemented using their biometric characteristics.

It is thus possible to use, in recognition systems, fingerprints, the iris of the eye, the ear, or even the whole head moφhology. But above all, we rely on the recognition of fingerprints.

However, this process, the most widespread, if it allows a high resolution, has many shortcomings.

Due to the large amount of information it can deliver, it is cumbersome and slow to implement.

Capturing fingerprints involves placing the underside of a finger on a grasping surface: - grease or perspiration covering the finger can alter the grasping;

- contact between the finger and the gripping surface can cause distortions;

- It is relatively easy to cover your finger with a false impression film, especially since it is hidden from the view of an operator.

In short, the plaintiff has sought a more flexible and faster alternative solution, if not more secure, and this is how she proposes her invention.

The invention relates to a method of identifying persons using biometric characteristics, according to which biometric characteristics of a person to be identified are entered, identification data are extracted therefrom and the identification data of the person compared to identify with reference data from a base, characterized in that an image of folds of the skin of the person to be identified is captured, the reference fold data having been previously acquired and stored in the base, and captures an image of the skin folds and performs a first recognition at by means of said image as a filtering step, before filtering reference data from the base relating to other biometric characteristics and continuing the recognition by means of these said other characteristics.

As folds used to identify people, we can consider, for example:

- the folds of the finger joints, preferably on the top of the hand, at the knuckles, - the wrinkles on the forehead.

In addition to the objectives aimed at the origin of the invention, recognition by means of skin folds offers the advantage of providing contextual information, that is to say presentable in alphanumeric form, on certain related characteristics of people. , as their size, their age, their sex, in any case more reliable than those which could be drawn from conventional biometric data (fingerprints, iris ...). Thus, if we find that the folds of an area of the skin of the hand are close enough to each other, we will deduce that it is probably a child rather than an adult.

In addition, and above all, by virtue of the method of the invention, it is possible to "cut out" the area of the base containing identification data of fingerprints, which is very dense, and to access the precise information sought more quickly.

Thanks to this recognition in several stages, the process of the invention is clearly distinguished from that which is taught by the document Joshi DG et al: "Computer vision-based approach to personal identification using finger crease pattern", Pattern recognition Perganion Press Inc. Elmsford, NY US, vol. 31, No. 1, 1998, p. 15-22.

Note, however, that in many applications, more rudimentary recognition by means of the folds alone will be sufficient while providing considerable advantages as regards the matching of the characteristics entered and the reference data in the database:

- by providing information that is rich and sufficiently stable over time to feed a recognition system, - by providing additional information to increase accuracy and speed.

The invention also relates to a person identification system, comprising at least one image sensor, an image processing module of the sensor to extract contextual information and biometric characteristics from it, a reference base containing contextual information. of reference and reference biometric characteristics, a block of comparison of the extracted contextual information and of reference context information and a block of matching of the extracted biometric characteristics and of the reference biometric characteristics, the block of comparison of contextual information providing indications for filtering the reference biometric characteristics to be matched.

Preferably, the image processing module includes a block for calculating contextual information linked to related characteristics of people and as contextual information, we will choose:

- the number of folds per joint and per finger, the distance between the fold zones of the various joints, and the length of these folds in a predetermined order of enumeration,

- the number of forehead wrinkles, their spacing and their length.

If the contextual information is sufficient to identify the person, the biometric characteristics cease to be used, the matching block being short-circuited.

Still preferably, a support surface of the hand of the persons to be identified is provided, determined relative to the focal plane of the image sensor and a complementary biometric sensor, for example a fingerprint sensor, integrated into said surface. , the reference biometric characteristics of the fingerprints being previously introduced into the database.

This version is particularly interesting in criminal identification applications.

Advantageously also, the pairing module can be connected to a memory card by means of a card reader equipping the system. The memory card then serves as a reference base for an authentication system.

Advantageously also, the pairing module is connected to a microphone fitted to the system. The reference base then contains reference voice data and the system is well suited as a doorman.

It will also be noted that the acquisition of fold data is less aggressive towards the person to be checked than that, for example, of fingerprints or of the iris, which is very appreciable, in particular in strictly civil applications.

Likewise, the identification process based on folds is more difficult to falsify, the detection of a false hand being able to be detected more easily than a false impression or a lens for the iris.

The invention will be better understood with the aid of the following description and of the appended drawing in which: FIG. 1 represents a diagram by functional blocks of the identification system according to the invention;

- Figure 2 is an image of the skin of the back of a hand;

- Figure 3 shows a block diagram of an authentication system implementing the method of the invention; - Figure 4 is a functional flow diagram implementing the method of the invention;

- Figure 5 illustrates the implementation of the system of the invention;

- Figure 6 is an alternative embodiment of Figure 5, with a fingerprint sensor; - Figure 7 is another alternative implementation of the system for capturing forehead wrinkles;

- Figure 8 is an exemplary embodiment of the system of the invention for an authentication implementation;

FIG. 9 is an alternative embodiment of the system of FIG. 8, for capturing forehead wrinkles and

FIG. 10 is an alternative embodiment of the system of FIG. 9.

With reference to FIG. 1, the system 1 comprises an image sensor 20, here a digital video camera delivering images, controlled by a presence detector 21. It can be an optical device composed of a light-emitting diode and a photodiode, or a pressure sensor, detecting the presence of a hand 10 of a person to be identified , on a support surface 11 of the hand allowing correct focusing.

The camera 20 is connected to an image processing module 30 which extracts the identification data of the images from the image sensor 20, and which is connected to a module 50 for matching the identification data from different sources, one of the sources being the sensor 20 and another, a reference base 40. The pairing module 50 is also connected to the base 40 and controls a module 60 for displaying the results of the identification, the latter being able to possibly order actions accordingly.

More specifically, the image processing module 30 includes a block 33 for converting images into identification data, and a block 32 for extracting biometric characteristics from the identification data, the function of which is also implemented when these are fingerprints. The characteristics then consist of a synthetic description of the drawings of the imprints (arches, loops, whorls, etc.).

Here, an image of the skin of the back of the hand 10 is captured. With reference to FIG. 2, the skin of the fingers appears striated with folds, especially at the joints, in particular between the phalanx 13 and the phalangin 14.

The identification data produced by the conversion block 33 are also transmitted to a block 31 for calculating contextual information from these identification data.

The contextual information is for example the number N of folds on each joint, the distance D separating them from the base of the nail, the transverse length L of the folds, possibly their inclination I on the longitudinal axis of the finger, or the number of B forks on the finger.

This information is of alphanumeric type and some of it is closely related to related characteristics of the people: the size, the age, the sex, so that being in the presence of a combination of these makes it possible to exclude the possibility of being in the presence of a combination of these and consequently to filter the corresponding biometric characteristics.

The base 40 is managed by an interface 43 connected to the pairing module

50 and comprises a memory 41 for reference contextual information and a memory 42 for reference biometric characteristics. The memories 41 and 42 have been divided into smaller memory zones 41i and 42i, which correspond, each zone of index i containing either a combination of contextual information (41i) or the reference biometric characteristics (42i), and both linked to a predetermined combination of the related characteristics mentioned above and designated by the index i. It is thus possible to select the biometric characteristics of an area 42i of the base from the recognition of the contextual information of an area 41 i of the base thanks to the index i which can serve as an indication of filtering.

The matching module 50 comprises a matching block 52 (matcher), known to those skilled in the art, capable of searching in the base 42 for the biometric reference characteristics closest to those extracted from the recorded image. on the person to be identified, calculate the probability with which they come from the same person, and deduce the result to be displayed.

The matching module 50 additionally includes a comparison block

51 which receives the contextual information calculated by the calculation block 31.

The blocks 51 and 52 are connected to the base 40 via the interface 43 and, at the output, to the display module 60.

If we want to achieve a system of authentication of people, and no longer of identification, with reference to FIG. 2, the system 2 includes all the elements of system 1 except that the base 40 is replaced by a memory card, or chip, 81 connected to the system by means of a card reader 82 containing the interface 43. The reader 82 additionally comprises a detector for the presence of the card which replaces the detector 21. In a simple version, the memory card contains a single zone 41i containing the contextual information of the card holder, the pairing module contains only the comparison block 51 and the processing module 30 does not need to contain the extraction block 32.

For more complex applications involving a combination of biometric sensors, as will be seen below, the two zones 41i and 42i, as well as the blocks 32 and 52, must be present.

The devices making it possible to capture the images of the skin under good conditions will now be described.

The system 1 comprises, in a simple version, with reference to FIG. 5, a bearing surface 11 for placing the hand 10, the sensor 20 being disposed above the surface 11 for correct focusing on the back of the hand .

In a more complex version, with reference to FIG. 6, involving a complementary fingerprint sensor 20e, integrated into the support surface 11 and comprising a pressure detector so that, by pressing on the sensor fingerprints, the camera is triggered, it then captures the image when the hand is correctly positioned.

In this latest version, the fingerprint identification data is attached to the skin identification data, both for the data entered and for the reference data. Instead of the folds of the skin of the hand, one can also capture, with reference to FIG. 7, the images of the wrinkles on the forehead 12, with a system whose same references designate the same elements.

The system 2 comprises, in a simple version, with reference to FIG. 8, a reader 82 for receiving and reading the card 81 presented by the hand 10. The sensor 20 is placed above the card inserted in the reader and can thus capture a picture of the back of the hand in good conditions. But we can also, with reference to Figure 9, capture an image of forehead wrinkles. For these last two versions, a complementary fingerprint sensor 20e can be located on the card itself, the authentication system 2 then having the same structure as the identification system 1.

The system 2, with reference to FIG. 10, can still find its application in devices of the residential door entry type, the complementary sensor here is a 20m microphone which can be integrated into a screen 11, here transparent, protecting the camera 20 , so that by pronouncing his name in front of the microphone, a person triggers the capture of the image by the camera, the forehead being suitably positioned. The voice data is attached to the identification data of the forehead skin, both for the data entered and for the reference data. This voice data is for example of the type provided by a frequency analysis of the person's speech.

The operation of the system will now be explained.

With reference to FIG. 4, when the camera 20 is triggered by the detector 21, it captures an image of the folds of the skin of the hand 10 or of the forehead 12 of the person to be identified. During a step 100, the image is transmitted to the conversion block 33 which converts the image into identification data of the person to be identified. Then, during a step 102, from these identification data, the extraction block 32 extracts bionietric characteristics of the image and, in parallel, from these same data, during a step 101, the calculation block 31 calculates the contextual information of the image.

The biometric characteristics extracted from the image could then, possibly in a step 105 executed by the matching block 52, be directly compared to the reference biometric characteristics of the memory 42 of the base 40. We would then proceed to display the results in a step 106 executed by the display module 60.

But the matching block should then scan all of memory 42, to compare the extracted biometric characteristics with all the reference characteristics stored in the base. Here, this disadvantage is avoided thanks to the possession of contextual information and to preliminary comparison 103 and initialization 104 steps of the interface 43.

During step 103, the comparison block 51 compares the contextual information provided by the calculation block 31 with all the contextual reference information of the memory 41, and retains only the closest combinations. These combinations are stored in the areas 41i of the memory 41. The indices i constitute the indications for filtering the reference biometric characteristics and are stored in a memory not shown of the interface 43. These operations, relating to alphanumeric data and not very large, are fast.

During step 104, the results of step 103 are used:

- if no index i or all the indices i are proposed, the filtering fails and, according to an option previously taken, either we go to step 105 already described to extend the complete search or we go to step 106 executed by the display module 60 to refuse identification;

if the index i is unique, it designates a unique area 42i and, if this area contains the reference biometric characteristics of only one person, the person is identified and the signaling step 106 directly provides his authentication, the matching block 52 being short-circuited;

- If one or more indices i are selected designating one or more zones 42i and at the same time the reference biometric characteristics contained therein, the other zones of the base 40 are separated by the interface 43. Step 104 then gives control of operations in step 105.

During the execution of a step 105 following the step 104, the identification is continued using only the selected biometric characteristics and the result is communicated to the display step 106.

Unlike the execution of step 105 not following step 104, the pairing block 52 scans only a few areas 42i of the memory 42 and the time saved is considerable.

Claims

1- Method of identifying people by means of biometric characteristics according to which one captures (20) biometric characteristics of a person to be identified, one extracts (100) identification data therefrom and compares (105) the data d identification of the person to be identified with reference data from a base (40), characterized in that an image of skin folds (10, 12) of the person to be identified is captured, the fold data having been previously acquired and stored in the base, and a first recognition is carried out by means of said image as a filtering step (103), before filtering (104) reference data from the base relating to other biometric characteristics and continue the recognition (105) by means of these said other characteristics.

2- The method of claim 1 wherein the folds used to identify people are selected from the group of folds of the finger joints, preferably on the top of the hand (10), at the joint of the phalanges, and forehead wrinkles (12).

3- Method according to one of claims 1 and 2 wherein the skin folds provide contextual information (N, D, L, I) presentable in alphanumeric form.

4- People identification system, comprising at least one image sensor (20), a sensor image processing module (30) for extracting (31,32) contextual information and biometric characteristics, a reference base (40, 41, 42) containing contextual reference information and reference biometric characteristics, a comparison block (51) of the extracted contextual information and reference contextual information and a matching block (52) of the extracted biometric characteristics and reference biometric characteristics, the block (51) for comparing contextual information providing indications for filtering the reference biometric characteristics to be matched. 5- The system of claim 4, wherein the image processing module (30) comprises a block (31) for calculating contextual data (N, D, L, I) linked to related characteristics of the person to be identified .

6- System according to one of claims 4 and 5, in which the contextual data are the number (N), the position (D) and the length (L) of the folds of the skin, or the number of forehead wrinkles, their spacing and length.

7- System according to one of claims 4 to 6, wherein the matching block (52) is short-circuited if the contextual information is sufficient to identify the staff.

8- System according to one of claims 4 to 7 wherein a fingerprint sensor (20e) is integrated into a surface (11) of the hand support.

9- System according to one of claims 4 to 7, wherein the pairing module (50) is connected to a memory card (81) via a card reader (82) and the memory of the card serves as a reference base (40).

10- The system of claim 9, wherein the reader (82) is arranged to trigger the image sensor (20).

11- System according to one of claims 4 to 7, wherein the pairing module (50) is connected to a microphone (20m).

12- The system of claim 11, wherein the microphone is arranged to trigger the image sensor (20).