DRAFT Study on biometrics in US Patent Literature.

 

By : Zeno Geradts

Date: 23 December 2002.

 

Introduction

 

For the Netherlands Forensic Science Institute a study has been carried out for information in US patent literature on developments in biometrics.

 

For this research all patent applications and patents have been searched with US Patent Code 382/115-118 that cover biometrics devices for speech, handwriting, finger and hand-scans.

 

The survey has been limited to US Patents, since examination time is limited to several days. A paper more in depth on the forensic aspects of biometric devices is available at http://forensic.to/biometrie.pdf.

 

For the patent codes, we can see the following statistics for the number of patents in the database until 15 December 2002(which are classified by the US Codes):

For the fingerprint section, 381 patents were retrieved.

 

The examination of patents in more depth is limited to those that are relevant and describe a new way of use of biometrics. Furthermore attention is given on prevention of tampering with biometrics device. The patent survey is limited to 1997-2002 and includes the applications of patents.


 

Finger and palm prints

 

Company-analysis

 

The assignees in the group of patents that have been examined were often small companies or individual inventors. Well-known companies as Toshiba, NEC, IBM, Sagem, Authentec and Digital Persona are included. A remarkable patent application is from the United States of America – the National Security Agency.

 

The company Authentec, which is a part of the Harris Corporation is specialized in OEM-products. The company Digital Persona has eight patents in this field and is focused on integration of finger print scanners. NEC is more focused on biometrics with smartcards, however there are also interesting patents in palm prints.

 

  

General

 

Several companies describe a total solution with encryption of the biometric data. Most card systems do not work with matching on the chip card, but with a separate computer (since processing power is limited on these chip cards).

 

Scanners

 

The company Advanced Precision Technology in California has issued several patents on finger scanners by means of a laser or a Spatial Light Modulator.

 

Patent US5920640 describes a fingerprint scanner of Harris in which live detection is integrated.

 

Patent US6091838 of E.L. Specialists describes a cheap fingerprint sensor that is very compact. A system with electrodes, which is also cheap to manufacture, is shown in US56940525 by Toshiba and by NEC (US6154580).

 

Fingerprint scanner by NEC US6154580

 

The design of the fingerprint scanner is a separate US class. An example of a special design is shown in a patent of Arete Association WO20000039743 for making a reproducible image of a finger. Another example is US6038322 (Digital Biometrics) in which a palm imaging system is described.

 

Algorithms

 

For fingerprint matching the next methods are available in patent literature

·        Minutiae extraction (which costs much computing power)

·        Direct image correlation (can also be implemented in ASICS hardware as is described in US5056162). A disadvantage is that a higher accuracy on calculation slows down the computing speed.

·        A sophisticated algorithm has been described in US6134340 by TRW-California. In this system pre-selection steps have been made for the shapes that are available in the fingerprints.

In this search also European Patents have been taken into account.

 

Fujitsu improves the reliability at the enrollment process by detecting the number of minutiae.

 

In EP1093045 Fujitisi describes and algorithm. Patent EP12033344 from Precise Biometrics with Veridicom describes a method for comparing templates. This algorithm is even better described in US6241288. The company has submitted this patent in many countries. Another patent application is EP1150608 for comparing partial fingerprints. A patent of the NSA in my opinion, anticipates this patent.

 

NEC has applied for a patent EP1063605 for registration of fingerprints. This is meant to compensate for rotations and translation. Casio describes a correlation algorithm in EP1017008.

 

Siemens describes in US5920641 a linear structure for fingerprints. A quality measure is used to evaluate the reliability of the analyzed data. Singularities and Minutiae are extracted and stored.

 

Auto calibration with a fingerprint scanner is described in EP1208524 by Cross Matching Techniques (this techniques is widely known from document scanners).

 

In US2002/0108047 (Bacon and Thomas) a patent is described for coping with different types of fingerprints scanners and comparing those images.

 

US2002/0154973 (Freedman & Associates) takes care of biometric data that changes in time. In the patent is also information available that users of biometric fingerprint access control systems will learn how to use the system. For that reason the more experienced user can have a higher level of verification.

 

Authentec (US6021211) describes an efficient method for indexing and searching in a plurality of reference fingerprints.

 

NEC describes several patents (e.g. US6370263) on palm prints and the registration of the several parts of the prints.

 

IBM discloses a patent in US5884971 and US6005963 with an exact algorithm for comparing fingerprints. They claim that their invention can cope with smudged impressions.

Algorithm of finger print matching described in US5884971

 

 

Power spectral density data in combination with normalized spatial correlation values is used in US2001/0016055  (Japanese application).

Applications

A number of applications have been found in patent literature. Many of those or focused on phones, ATM’s, firearms and airports.

Firearm holster with finger print identification from US6320975.

Live detection and tampering control

 

In US2002/0095608 a switch is described against tampering with the electronic device. It is isolated from its power source; it is not possible to circumvent the access control feature.

 

In US2002/0122571 a method is described against fraudulent copying of the digital stored biometrics information.

 

In US2002/0138768 a portable system is described for examining the biometric evidence further with a heartbeat waveform and by measurements taken by the reflecting light of sub dermal layers of the skin tissues.

 

In US 2002/0146154 a method is described by detecting tampering with the system if the biometric sample is exactly the same as the enrolment data. Normally there should be dynamic changes in the applicant’s biometric samples.

 

In US6373967 (California Institute of Technology) a system is described in which fingerprints must be entered in a proper sequence to be recognized by the system, which is known to the user.

 

In US5719950 (Minnesota Mining and Manufacture Company: 3M) a biometric authentication system is disclosed. In this system the stored biometric data is compared with one non-specific biometric parameter of a physiological characteristic. Claimed is that it is possible to determine that an individual is not incapacitated, dismembered or deceased. From the Internet http://www.ait.ca/html/solutions/biometrics.html it appears that 3M is active in Biometrics for integrated solutions.

3M patent US5719950 determination of physiological characteristics

 

In US5737439 (Smartouch) describes an anti-fraud biometric scanner that determines if an object exhibits the characteristics of a live human. It has a scanning means for obtaining the biometric sample from the object, blood flow detection. It is also determined if someone is trying to simulate blood flow. From the Internet not many products have been found from this company. One page that appears to be from this company is www.smartouch.net appears to be static.

 

US5987153 (Quintet) involves detection of fraud with biometric data such as signatures. If the signature is exactly the same as the stored signature, this is rejected. This patent anticipates US 2002/0146154. A link to Quintet hand writing verification can be found at http://www.quintetusa.com /

 

US2001/0031071 (Medtronic) describes access to medical devices (IMDs) by biometric data.

 

Example of access to medical devices by biometric data (US2001/0031071)

 

Several patents (e.g. US62598705 (Dew Engineering)) describe the use of encryption for improving the security of these systems.


 

Other methods with hands and fingers

US 5793881 describes a system with the structure and other characteristics of the hand by the way someone uses a door handle.

US5862246 shows a system with verification of the knuckles surface of one hand.

US 5787185 shows a patent of the British Technology Group with verification based on the vein pattern in a hand (as is shown in Figure 4 of this patent below). A Korean company BK Systems also describes the vein pattern in a hand. As is shown in http://www.soluniverse.com.hk/bksystem.htm this is also a commercial product.

Furthermore we have found a patent US5751835 of identification of individuals based on their nail patterns.
Faces

On face recognition many different institutes and companies have filed patents.

 

Patents are focused on recognizing faces are tracking persons, as can be seen in US6445810 from Interval Research Corporation. Some of these systems are focused on recognizing several persons in one image, as is shown in US6466685 from Toshiba. Face detection using templates is described in US5835616.

 

In US2001/0005222 a system is shown in which the skin pigmentation is taken into account.

 

A combination of speaker recognition and movement of the face is shown in US5625704 by Ricoh. The use of movement identification is also used in a patent of the Fraunhofer Institut in US6101264.

 

An even better approach is described in  US6492986   Method for human face shape and motion estimation based on integrating optical flow and deformable models.

 

Unisys shows in US6026188 a system where a three-dimensional face is stored with several enrollment images in several positions. A wire frame model is implemented by DEC in US6016148. The wire frame model is used for searching in the image database.

 

Of companies that are shown in the open directory project:

Identix has eight US Patents, however they are mostly focused on fingerprint matching.  Visionics has one patent of which one is focused on an application.

 

 

 

 

Detection of features in the face by visual image representation of mouth movements (US5625704)

 

 

A method against tampering with facial systems using photographs is described in US2002113687.

 

Philips describes an adaptive face recognition system in US2002136433 

 

Algorithms

 

Lucent describes face recognition using DCT’s in US5802208. The Instituto Trentino Di Cultura shows in patent 5764790 features that are extracted from a database of image and shows the exact algorithms that have been implemented. Mitsibushi shows a face image processing system capable of eye detection.

 

Mathushita (US6345109 ) describes a face recognition method that can cope with different image devices and levels of noise.
 

IRIS and retina Scans

 

US5291560 of Iris scan (Daughman) incorporated is the ground patent in the field of iris scans. It has been filed in 1991. However, identification by eyes is also described in US4641349 (1987) and iris identification is even described in 1977 : US4109237 (this patent is expired now).

 

Automatic detection from US4109237

 

Sensar (previous, Iridian Technology and Iris Scan) describes the

implementation of an iris scan in a cell phone (US5377699).

 

One company that has several patents for iris recognition on animals, e.g. horses is Oki in Japan (US6373968). Oki has licensed the patents of Sensar and uses it in ATMs http://www.iris-scan.com/iris_recognition_today.htm).

 

National Instruments Corporation (US 5353233 and US6434515) has a software patent on the algorithm that can be used (with Gabor filters) for comparing the iris (references are made to the patent of Daughman). The software source code is also available.

 

British Telecom (US6333988) describes an IRIS scan method and apparatus.

 

Sensar US6320610 and describes a patent with a compact imaging device, which can be used in ATMs. Another patent is shown in US6262977, which is used for iris scan through eyeglasses. (http://www.iridiantech.com/about.php?page=4) .

 

Another equipment is visualized in US6021210 to detect the IRIS in a face.

Patent US6021210 method of iris extraction by Sensar

 

Another company, Eyeticket, describes a system for ticket less aircraft check-in using iris scans in US6119096. The companies Eyeticket and Sensar have a strategic alliance.

 

The company of Sensar has more US patents, which might be relevant for product evaluation:

     US6332193   Method and apparatus for securely transmitting and authenticating biometric data over a network

     US6320610   Compact imaging device incorporating rotably mounted cameras

     US6299306   Method and apparatus for positioning subjects using a holographic optical element

     US6088470   Method and apparatus for removal of bright or dark spots by the fusion of multiple images

     US6064752   Method and apparatus for positioning subjects before a single camera

     US6021210   Image subtraction to remove ambient illumination

     US6005704   Cable driven image steering and focusing device

     US5978494   Method of selecting the best enroll image for personal identification

     US5953440   Method of measuring the focus of close-up images of eyes

     US5428357   High speed data acquisition system and method


 

Handwriting

 

US6381344 shows a computer-based method and system for capturing and verifying a handwritten signature.

 

US6445820 (Limbic Systems) describes a method for electronic analysis of handwriting samples by creating 3D representations of portions of the writing sample.

 

3D-image of Limbic Systems patent

 

US5818955 of Penop Limited (England) describes a method for signature verification.

 

US5745591 discusses a system for verifying the identity of a person while holding an object and/or performing a task such as writing. Electrical transducers are used for measuring this task.


 

Speaker recognition

 

Speaker recognition was outside the scope of this research. In the patents that we have retrieved only two were focused on speaker recognition, whereas we could easily find over 300 US Patent (Applications) on this subject.


 

Miscellaneous

 

US6205233 (Invisitech) describes a patent of sensors with weight, height, voice timbre and time to perform different functions.

 

Sensors shown in image of US6205233

 

Another patent with gait parameters is US2002/0028003 in which the anatomy / gait is recognized from a database.

 

US6421453 shows the possibilities in which user recognition is accomplished by behavioral passwords.

 

US5878155 describes the use of a bar code or a design that is tattooed on an individual for identification purposes.

 

In US5787187 of Sandia Corporation the acoustic properties of the ear channel are used as a biometric measure.

 

US6038465 (Agilent Technologies – a Telemedicine company) discusses a system and method for identification of the user based on weight and by reflecting sound waves at locations that defines the length.

 

Example of patent 6421451 (IBM) of behavioral passwords

 

Finally US6181804 describes a method for comparing shoeprints with a computer. These claims in this patent are too broad, and this might have been caused by the fact that not much patent literature is available, however it is a general practice to do a comparison like this, as can be seen in forensic literature.