WO2006082550A1

WO2006082550A1 - Biometric identification apparatus using fluorescence spectroscopy

Info

Publication number: WO2006082550A1
Application number: PCT/IB2006/050305
Authority: WO
Inventors: Maarten M. J. W. Van Herpen; Cristian Presura
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2005-02-07
Filing date: 2006-01-27
Publication date: 2006-08-10

Abstract

A biometric identification apparatus (2) and accompanying method is disclosed that uses fluorescence spectroscopy as a validation means for use in identification. According to the invention, a particular embodiment comprises an image capture means (4) arranged to sample an object (18) and to create an electrical representation of this object. This electrical representation is passed on to an image matching means (6) that matches said electrical representation with entries in a reference database (8). The apparatus further comprises a spectroscopic analysis means (12) arranged to measure spectroscopic features of the object using fluorescence spectroscopy. These spectroscopic features are processed by a spectroscopic matching means (14) that matches them with reference spectroscopic features. Finally, a decision means (10) combines the information from multiple sources to a decision (16).

Description

BIOMETRIC IDENTIFICATION APPARATUS USING FLUORESCENCE SPECTROSCOPY

The invention relates to a biometric identification apparatus comprising an image capture means arranged to create an electrical representation of said object, and an image matching means arranged to match said electrical representation with entries in a reference database.

Biometric identification apparatus such as fingerprint identification apparatus are widely used. It was shown by Tsutomu Matsumoto, et al. in "Impact of Artificial "Gummy" Fingers on Fingerprint Systems" (Proceedings of SPIE Vol. #4677, Optical Security and Counterfeit Deterrence Techniques IV, Thursday-Friday 24-25 January 2002) that many fingerprint identification apparatus could be fooled by the use of a "gummy" finger.

To frustrate spoof attempts using "gummy" fingers, fingerprint identification apparatus can be combined with other means that can differentiate between "gummy" fingers and human fingers.

A biometric identification apparatus that can be used to detect such a spoof is described in U.S. Patent No. 6,816,605 entitled "Methods and systems for biometrics identification of individuals using linear optical spectroscopy."

The biometric identification apparatus as disclosed in U.S. Patent No. 6,816,605 uses linear spectroscopy to analyze electromagnetic radiation scattered by tissue from an individual. The measured spectral variation is compared with pre-recorded spectral variations over a predetermined wavelength interval. The individual is designated as having an identity associated with the pre-recorded spectral variation if the measured spectral variation is consistent with the pre-recorded spectral variation. An apparatus as disclosed in U.S. Patent No. 6,816,605 can be combined with a fingerprint identification apparatus, as suggested in the aforementioned U.S. Patent. The resulting system can detect spoof attempts that use a "gummy" finger. The aforementioned system has several drawbacks; it requires the enrolment of a plurality of spectral variations of living tissue, and it requires multivariate analysis to establish liveness.

It is an object of the invention to provide an alternate means for validating the outcome of a biometric identification apparatus.

This object is realized in that the biometric identification apparatus of the type set forth in the opening paragraph further comprises: - a spectroscopic analysis means arranged to measure spectroscopic features of said object using fluorescence spectroscopy, a spectroscopic matching means arranged to match said spectroscopic features with reference spectroscopic features, a decision means arranged to combine information into a decision, said information comprising information from the image matching means and information from the spectroscopic matching means, said decision identifying whether the object sufficiently matches an entry in the reference database.

The spectroscopic features measured by the spectroscopic analysis means are passed to the spectroscopic matching means where they are matched with spectroscopic features of entries in the reference database. This match may take place on spectroscopic features of individual entries, or spectroscopic features that hold for classes of entries. When the measured features match within a predefined tolerance, there is a spectroscopic match.

In the decision means, information from the image matching means as well as the spectroscopic matching means are combined, possibly with other information such as an alleged identity, to a decision. This decision comprises a positive or negative identification. In the case of a positive identification, this decision may further comprise the established identity of the object and/or a reliability measure.

Fluorescence spectroscopy or fluorometry is a type of electromagnetic spectroscopy used for analyzing fluorescence spectra. In general, it involves directing electromagnetic radiation, such as ultraviolet light, at an object. The electromagnetic radiation can excite electrons in molecules in the object. When these electrons fall back to a lesser or non-excited state, they emit electromagnetic radiation with an energy content lower than, or equal to, that of the original exciting electromagnetic energy. The spectrum of the emitted electromagnetic radiation can be measured by using a spectrometer. The emitted radiation comprises information about the energy of molecular vibrations and rotations; these in turn depend, inter alia, on the particular atoms or ions that comprise the molecule, the chemical bonds that connect them, and the symmetry of their molecule structure. In embodiments of the invention, fluorescence spectroscopy is used to obtain information on the presence of characteristic fluorophores in the object offered for identification. This information can be combined with information from the image matching means and can help in discovering spoof attempts making use of replicas that do not share spectroscopic features characteristic of the objects being identified. Human skin tissue comprises native fluorophores, such as collagen, and tryptophan. Embodiments of the invention use fluorescence spectroscopy to indicate the presence of such fluorophores in the object offered for identification. For example, in embodiments that focus on fingerprinting, fluorescence spectroscopy can be used to determine whether or not the object offered for identification exhibits spectroscopic features characteristic of skin tissue on human fingers. When a "gummy" finger is applied to the identification means, it may pass the image matching means, but it will not pass the spectroscopic matching means as it does not comprise fluorophores such as tryptophan, or collagen characteristic of human skin.

In embodiments that determine whether the object offered for identification has particular spectroscopic features common to all entries in the reference database, there is no need for enrolment of spectroscopic features for each individual entry in the reference database.

A particular advantage of fluorescence spectroscopy results from the fact that there is a lag between the excitation of electrons in the object by incoming electromagnetic radiation and the radiation emitted by the object. When the radiation source is switched off, the object will continue to emit radiation for a short period of time. During this period, there will be little background radiation, and sensitive equipment can be used to measure the emission spectrum. As a result spectroscopic feature acquisition is simplified. Furthermore the present invention does nor require complex multi-variate analysis to validate the identification outcome.

It is a further object of the invention to provide an alternate method of validating the outcome of a biometric identification method.

This object is realized by a biometric identification method comprising: creating an electrical representation of an object, a first matching comprising matching said electrical representation with entries in a reference database, the method further comprising: measuring spectroscopic features of said object (18) using fluorescence spectroscopy, a second matching comprising matching said spectroscopic features with reference spectroscopic features, and combining information into a decision, said information comprising information from the first and the second matching, said decision identifying whether the object(18) sufficiently matches an entry in the reference database (8).

These and other aspects of the biometric identification apparatus will be further elucidated and described with reference to the drawings, in which:

Fig. 1 is a block diagram of an embodiment of a biometric identification apparatus for one-to-many identification according to the invention;

Fig. 2 is a block diagram of an embodiment of a biometric identification apparatus for one-to-one identification according to the invention; Fig. 3 is a schematic overview of an image capture means for use in a biometric identification apparatus according to the invention;

Fig. 4 is a schematic overview of a spectroscopic analysis means for use in a biometric identification apparatus according to the invention;

Fig. 5 is an alternative schematic overview of a spectroscopic analysis means for use in a biometric identification apparatus according to the invention;

Fig. 6 is a schematic overview of a combined image capture means and spectroscopic analysis means for use in a biometric identification apparatus according to the invention.

Throughout the drawings, identical reference numerals refer to the same elements, or to elements that perform the same function.

There are two important classes of biometric identification: one-to-many identification and one-to-one identification. The invention can be used for both classes. In one-to-many identification, an object is presented for identification and is matched with a plurality of entries contained in a reference database. A positive identification occurs when a match is found within a predefined tolerance. A negative identification occurs when no such match can be made. In certain embodiments, it is sufficient to decide on either a positive or a negative identification. One application may be a system that uses a biometric identification apparatus for access control. Other applications, such as information systems for security personnel, call for other embodiments in which there is also a need to establish the identity and or a reliability measure for the identification.

In one-to-one identification, the biometric identification apparatus additionally requires an alleged identity. In the case of one-to-one identification, a positive identification occurs when a sufficient match is found between the measured information and the information in the reference database associated with the alleged identity. A negative identification occurs when this is not the case. In certain embodiments that focus on one-to- one identification, it is useful to additionally establish a reliability measure. Both types of biometric identification apparatus will be further explained.

Fig. 1 is a block diagram of an embodiment of a biometric identification apparatus 2 according to the invention, suitable for one-to-many identification. It further shows a possible partitioning of the apparatus into sub-modules. The apparatus in Fig. 1 comprises the following sub-modules: - an image capture means 4, an image matching means 6, a reference database 8, a spectroscopic analysis means 12, a spectroscopic matching means 14, and - a decision means 10.

The image capture means 4 samples an object 18 and creates an electrical representation of this object. This representation is sent to the image matching means 6. In parallel, the spectroscopic analysis means 12 can perform a spectroscopic analysis of the fluorescence spectrum of the object 18. The spectroscopic features measured by this means are then sent to the spectroscopic matching means 14.

Finally, the decision means 10 comprises an identity means 28 for establishing the object's identity, and a reliability means 30 for establishing a reliability measure for the decision. The decision means 10 uses information from multiple sources to come to a decision 16 that comprises: - a positive or a negative identification, the object identity established by the identity means 28, and a measure of reliability, established by the reliability means 30.

In the one-to-many embodiment shown in Fig. 1, the image matching means 6 comprises a selection means 24. The selection means will use the electrical representation captured by the image capture means 4 to find the electrical representation in the reference database 8 that provides the best match. In Fig. 1, a best match verification means 26 subsequently determines whether the best match is within a predefined tolerance. Other refined embodiments can be envisaged in which the reference database 8 is extended with additional data structures to speed up the selection means 24. Both the selection means 24 and the best match verification means 26 can be implemented by using dedicated hardware or by using instructions for a general-purpose processor. Alternatively, a mix of both can also be envisaged.

Embodiments of the invention may vary from simple to complex. An example of a simple embodiment would be a one-to-many fingerprint identification apparatus that uses a fluorescence spectroscopy to establish the presence of tryptophan in the object. In this embodiment, the reference database 8 will contain the electrical representation of fingerprints for each entry in the reference database 8, as well as a single set of reference spectroscopic features and tolerances representative of all entries in the reference database 8. The decision means 10 subsequently has to verify only whether the measured spectroscopic features are in line with these reference spectroscopic features so as to approve the result from the image matching means 6.

A more complex embodiment of a one-to-many fingerprint identification apparatus may make use of a reference database 8 that contains both electrical representations of fingerprints and spectroscopic features for each entry in the reference database. The apparatus first establishes an interim identity 22 in the image matching means 6. This is subsequently used to perform a match of the measured spectroscopic features with the spectroscopic features in the reference database 8 for this particular interim identity 22. The latter method is particularly effective as the spectroscopic features can, but need not, be unique for each individual. In addition, the tolerances used in this comparison only need to account for intra-person variations and need not be generalized over multiple persons.

When reliability is the key, the spectroscopic analysis means 12 can be further enhanced by adding additional data to the reference database 8. A reference database 8 may comprise: - spectroscopic features characteristic of each entry, tolerances for these features for each entry that allow selection based on these tolerances, relationships between spectroscopic features for each entry, and tolerances for these relationships for each entry that allow selection based on these relationships.

Fig. 2 is a block diagram of an embodiment of a biometric identification apparatus 3 according to the invention, suitable for one-to-one identification. The biometric identification apparatus 3 in Fig. 2 differs from that in Fig. 1 in that the image matching means comprises different building blocks.

In the embodiment shown in Fig. 2, the image matching means 7 requires an alleged identity 20. To this end, the image matching means 7 comprises a requesting means 32 that obtains an alleged identity 20 of the object. An implementation of such a requesting means may be a badge reader, or a machine reader for passports. The image matching means further comprises an identity verification means 36. The alleged identity 20 is used by the identity verification means 36 to determine whether a match can be made with entries in the reference database 8. This process can be implemented more efficiently by adding a determination means 34 to the image matching means 7 that first establishes whether a reference database entry is available for the alleged identity 20 and locates it.

When present, the identity verification means 36 can ascertain whether there is a sufficient match between the captured electrical representation and the representation associated with the alleged identity 20 in the reference database 8. The identity verification means 36 can be implemented by using dedicated hardware or by using instructions for a general-purpose processor. Alternatively, a mix of both can also be envisaged.

Fig. 3 is a schematic overview of an image capture means 4 for a fingerprint identification apparatus according to the invention. An object 40 (finger) presented for identification is placed on a transparent plate 42. On the left-hand side, a light source 44 is shown that outputs light in the visible spectrum. This light is focused onto the object surface by means of a lens 46. The focused light irradiates the object at a predefined angle.

According to Snell's law, only light reflected by the ridgelines of the fingerprint is fully reflected through the transparent plate (total internal reflection), and is subsequently focused onto an image capture device 50 (here a detector array) by means of a lens 48. The detector array samples the incoming light and transforms the image of ridgelines into an electrical representation 52.

The invention is not limited to fingerprint identification apparatus; other embodiments are conceivable, such as sweat duct-based identification apparatus. Here, the object being identified is no longer limited to a finger, but may equally well be the palm of a hand or another part of the human skin. In fact, the invention is not limited to biometric identification apparatus for humans. Alternative embodiments might include biometric identification apparatus for cattle or livestock.

Fig. 4 is a schematic representation of a spectroscopic analysis means 12 for a fingerprint-based identification apparatus according to the invention. An object 40 (finger) presented for analysis is positioned on a transparent plate 42. The spectroscopic analysis means comprises an electromagnetic radiation source 54, including a pre-filter 56, and an optical fiber 58 used to transport said radiation towards the skin tissue of the object 40.

A second optical fiber 62 is used to transport part of the radiation emitted by the object to a post-filter 64. Shielding 60 is provided around the top of the optical fibers to prevent stray radiation from disturbing the measurement. The output of the post-filter 64 is passed to a spectroscopic measurement means 66 to determine spectroscopic features of the light emitted by the object 40. The terms pre-filter and post-filter do not exclude the use of a tunable pre-filter or a tunable post-filter. Furthermore, not all embodiments require the pre- filter 56 or the post-filter 64. The electromagnetic radiation source 54 used in certain embodiments can be selected from a wide variety of existing radiation sources. Certain radiation sources may provide particular advantages. An Ultra Violet (UV) emitting Light Emitting Diode (LED), for example, can provide a relatively simple and low-cost radiation source. The use of multiple UV emitting LEDs allows the use of multiple excitation wavelengths at a reasonable cost. Alternatively, common available broadband radiation sources can be used, which may require a pre-filter to attenuate or filter undesirable spectral components in the excitation spectrum. Tunable lasers in turn can also provide multiple wavelengths but do not require the additional pre-filtering.

Apart from the factor of cost, there are other factors that determine the suitability of a particular radiation source design. One of the key factors is the type of fluorophores present in the object. Tryptophan, for example, is known to emit light with a wavelength of 360 nm, and a radiation source generating radiation of a shorter wavelength is needed to excite electrons in tryptophan.

The wavelength is equally important for another reason. The shorter the wavelength of electromagnetic radiation, the higher the energy content of the individual photons. Radiation of short wavelength UV is known to cause damage of the skin and, consequently, short wavelength UV is less suitable for application on living human skin.

The spectroscopic analysis means depicted in Fig. 4 comprises a spectroscopic measurement means 66. Various designs of spectroscopic measurement means for fluorescence spectroscopy are known in the art. Spectroscopic measurement means generally decompose the incoming radiation into spectral components by means of prisms, gratings or other means known in the art. These spectral components in turn are subsequently measured by means of a detector array. Embodiments of the current invention may comprise simple spectroscopic analysis means 12 that focus on the presence of a particular spectroscopic feature in the fluorescence spectrum after excitation with a single burst of a particular wavelength. Alternatively, more complex spectroscopic analysis means 12 can be deployed on the basis of customer requirements with respect to reliability and cost. Complex spectroscopic analysis may involve excitation using a burst of electromagnetic radiation, comprising a plurality of time shapes and wavelengths and subsequent measurement of a plurality of spectroscopic features in a wide spectral band.

In certain embodiments, it may be advantageous to use the pre-filter 56 for encoding a pattern in the excitation radiation, or filtering spectral components from the excitation radiation. Likewise, the use of the post-filter 64 can help to decode the pattern or filter spectral components outside the area of interest.

Fig. 5 depicts an alternative spectroscopic analysis means 12 for use in a biometric identification apparatus according to the invention. It uses a beam- splitting mirror 70 rather than optical fibers to direct light onto an object 40 (finger). The object 40 presented for analysis is positioned on a transparent plate 42. The spectroscopic analysis means comprises an electromagnetic radiation source 54 and the pre-filter 56. The radiation that passes through the pre-filter is incident on a beam- splitting mirror 70 that directs the radiation towards the object 40.

Part of the emitted electromagnetic radiation will pass through the beam- splitting mirror 70 towards the post-filter 64. In embodiments with a post-filter, it can attenuate or filter spectral components related to the exciting electromagnetic radiation; the remaining electromagnetic radiation is directed towards the spectroscopic measurement means 66. The resulting spectroscopic features 68 are sent to the spectroscopic matching means 14. The terms pre-filter and post-filter do not exclude the use of a tunable pre-filter or a tunable post-filter.

Fig. 6 is a schematic representation of a combined image capture means and spectroscopic analysis means. In this implementation, both the image capture means and the spectroscopic analysis means measure characteristics of the same object 40 (finger). They focus on different characteristics of the object and, in doing so, improve reliability in general and can frustrate spoof attempts.

The spectroscopic analysis means depicted in Fig. 6 has a construction which is identical to that in Fig. 4. In a similar fashion, the image capture means in Fig. 6 has a construction which is identical to that in Fig. 2.

Embodiments such as the one in Fig. 6, performing simultaneous image capture and spectroscopic analysis, can benefit from the pre-filter 56 and the post-filter 64. The pre-filter 56 can attenuate or filter unwanted spectral components that may disturb the image capture means or spectroscopic measurement means 66. The post-filter 64 can attenuate or filter spectral components that may disturb the spectroscopic measurement means 66 or are outside the scope of the analysis.

Other embodiments may combine the image capture means with the spectroscopic analysis means in a time-multiplexed fashion. As long as the individual measurements are not perceived as such by the end-user, such choices do not affect the overall reliability.

Claims

CLAIMS:

1. A biometric identification apparatus (2,3) comprising: an image capture means (4) arranged to create an electrical representation (52) of said object (18), and an image matching means (6,7) arranged to match said electrical representation (52) with entries in a reference database (8), the biometric identification apparatus (2,3) further comprising: a spectroscopic analysis means (12) arranged to measure spectroscopic features of said object (18) using fluorescence spectroscopy, a spectroscopic matching means (14) arranged to match said spectroscopic features with reference spectroscopic features, and a decision means (10) arranged to combine information into a decision, said information comprising information from the image matching means (6,7) and information from the spectroscopic matching means (14), said decision identifying whether the object (18) sufficiently matches an entry in the reference database (8).

2. A biometric identification apparatus (2,3) as claimed in claim 1, wherein the decision means (10) further comprises an identity means (28) for establishing the identity of said object (18).

3. A biometric identification apparatus (2,3) as claimed in claim 1, wherein the decision means (10) further comprises a reliability means (30) for establishing a measure of reliability for the decision.

4. A biometric identification apparatus (2,3) as claimed in claim 1, 2, or 3, wherein the spectroscopic analysis means (12) comprises an electromagnetic radiation source (54) and a spectroscopic measurement means (66) for measuring spectroscopic features of electromagnetic radiation emitted by said object (18) resulting from electromagnetic radiation directed at the object (18).

5. A biometric identification apparatus (2,3) as claimed in claim 4, wherein said electromagnetic radiation source (54) comprises a tunable laser arranged to send a burst of electromagnetic radiation of a predetermined time shape comprising radiation of a predetermined wavelength.

6. A biometric identification apparatus (2,3) as claimed in claim 4, wherein said electromagnetic radiation source (54) comprises a narrow-band electromagnetic radiation source to send a burst of electromagnetic radiation of a predetermined time shape.

7. A biometric identification apparatus as claimed in claim 4, wherein said electromagnetic radiation source (54) comprises a broadband electromagnetic radiation source arranged to send a burst of electromagnetic radiation of a predetermined time shape.

8. A biometric identification apparatus (2,3) as claimed in claim 1, 2, 3, or 4, wherein said image capture means (4) comprises an image capture device (50) arranged to sample ridgelines of the fingerprint of the object (18) and create an electrical representation (52) thereof.

9. A biometric identification apparatus (2,3) as claimed in claim 1, 2, 3, or 4, wherein said image capture means (4) comprises an image capture device (50) arranged to sample the position of sweat ducts in skin of the object (18) and create an electrical representation (52) thereof.

10. A biometric identification apparatus (2) as claimed in claim 1, 2, 3, or 4, wherein said image matching means (6) comprises: a selection means (24) to select the electrical representation in the reference database (8) that matches the electrical representation (52) captured by the image capture means (4) best, and a best match verification means (26) to verify that the difference between the electrical representation (52) captured by the image capture means (4) and the electrical representation of the best match are within a predefined tolerance.

11. A biometric identification apparatus (3) as claimed in claim 1, 2, 3, or 4, wherein said image matching means (7) comprises: a requesting means (32) to obtain an alleged identity, and an identity verification means (36) to verify that the electrical representation (52) captured by the image capture means (4) matches, within a predefined tolerance, with the electrical representation stored in the reference database (8) for the alleged identity.

12. A biometric identification apparatus (3) as claimed in claim 11, wherein said image matching means (7) comprises a determination means (34) that determines whether an electrical representation for the alleged identity is in the reference database (8).

13. A biometric identification apparatus (3) as claimed in claim 1, 2, 3, or 4, arranged to measure spectroscopic features of electromagnetic radiation emitted by said object (18) characteristic of tryptophan.

14. A biometric identification apparatus (3) as claimed in claim 1, 2, 3, or 4, arranged to measure spectroscopic features of electromagnetic radiation emitted by said object (18) characteristic of collagen.

15. A biometric identification method comprising: creating an electrical representation (52) of an object (18), - a first matching comprising matching said electrical representation (52) with entries in a reference database (8), the method further comprising: measuring spectroscopic features of said object (18) using fluorescence spectroscopy, - a second matching comprising matching said spectroscopic features with reference spectroscopic features, and combining information into a decision, said information comprising information from the first and the second matching, said decision identifying whether the object (18) sufficiently matches an entry in the reference database (8).