Hand Geometry Identification System

 

EEL 4011 Senior Design II

Spring 2000

Luis F. Majano

Professor: Dr. Malek Adjouadi

Department of Electrical and Computer Engineering

Florida International University

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Project Justification

“Biometric technology holds limitless promise in the fields of personal identification and security.  The ability to implement different levels of security based on the accuracy of a submission or the value of the item protected is simply unthinkable in any password or key-based setting.   As biometrics makes inroads into the public realm, the layman's frequent reaction is hesitation, even trepidation; but what's so scary about biometrics?  In a word, privacy - or the lack thereof.” By IBG Congressional Testimony. 

           

The target of this project is to create a unique identification system.  Biometrical identification has been around now for quite some time.  However, none of the biometrical identification systems on the market have a low error percentage, when it comes to identification, and an affordable price.  We therefore decided to build a “LINEAR” hand scanning system.  The system will be investigated later on.

 

Sandia National Laboratories is a national security laboratory, which have been doing research on all biometrical identification systems.  They did a report on the evaluation of six different biometrical devices:

 

1)      Hand Geometry by Recognition Systems, INC.

2)      Fingerprint by Identix, INC

3)      Signature Recognition, Capital Security Systems, INC

4)      Retinal Scan by Eyedentify, INC

5)      Voice recognition by Alpha Microsystems, INC

6)      Voice recognition by International Electronics, INC

 

 

 

 

 

 

 

The following is a table of results that were encountered by Sandia National Laboratories.

 

Device

Accepted Response

Rejected Response

Acceptance Ratio

Voice (Alpha)

22

103

0.21

Voice (Int.Elec)

22

161

0.14

Retinal Scan

87

87

1.00

Fingerprint

60

78

0.77

Signature

22

69

0.31

Hand Geometry

181

11

16.45

 

The table clearly shows how the hand geometry system achieves a better performance above all of the systems in the recognition of the user.  It shows that out of 16.45 users, 1 was rejected.  Our system uses the same concept of the hand geometry system.  However, our system is a linear scanner of the hand.  This provides a more cost-effective system with the same acceptance ratio.  In response to the question "What is the best biometric?" one must immediately qualify the application in mind. For different applications, the ideal biometric will vary. An extreme example will illustrate the point: If you are accessing a nuclear facility, the effort and the intrusiveness related to the identification technique is not of consequence. If the scientist entering the facility needs to spend ten minutes using the system, so be it; if they feel offended by the technique or feels it invades their privacy, they can find a new job. At the other extreme, if a biometric is needed to allow an annual ticket holder into an amusement park, the worse case is to make the user feel at all uncomfortable or waste their time.  This is a basis of our design.  We want a product directed towards the average consumer or business that would want a low-cost, but reliable, identification system.

 

Please look at the graph below:

 

All biometrical devices have an acceptance ratio, which determines the scanning error for the device, by which a user can be accepted.  This ratio can be viewed as a percentage from 0 to 100%.  If the device has a 0% error, then the device will refuse entry to an intruder 100% of the time; false acceptance.  However, there is a trade off.  We have to consider the other important aspect of a biometrical device and that is the False Rejection.

The graph below depicts the ideal biometrical system and all of the now available systems.  We can observe that hand geometry systems have fewer tradeoff in response to accuracy, intrusiveness, cost, and effort.  Meanwhile, the other biometrical systems have these characteristics spread out.

 

System Criteria

The security requirements vary by application as well. In the case of the nuclear facility, a very high level of security is needed, whereas with the amusement park, a lower level is required. Accuracy is comprised of multiple parts: false acceptance (an imposter is accepted), false rejection (a legitimate person is rejected, a.k.a. insult rate), and non-acceptance (the system is unable to process a user, e.g., a fingerprint system with a user with no fingers). Likewise, in each case, the amount of money that can be afforded to building each system varies tremendously. The nuclear facility may have only two entrances, each of which can have a million dollars budgeted to protect access. The amusement part may have hundreds of entrances with a much smaller dollar per entrance budget.

            If we have a 0% false acceptance, then we have a 100% false rejection.  In this system, a legitimate user will be rejected 100% of the time.  With these characteristics in view, we can achieve a system, which approximates equilibrium on both sides of the spectrum.  Look at the chart below:

 

Optimal Point
 
 

 


This chart also depicts how the detection algorithm of the system can be coded.  We may move either to the up in the chart and obtain more false acceptance of an intruder, but lower our rejection of a normal user.  Or we can move down and lower the entry of an intruder, but have a higher rejection of legal users.  Therefore the optimal point is where we have decided to base our detection algorithm on.

 

 

 

Justification of Our Linear Scanner.

            In order for us to get a broader idea of what we where getting into, we decided to do a statistical report on the finger measurements of several individuals.  To do this, we created five strategically pegs on a transparency and decided to do 10 photocopies of each individual’s hand.  From these results we tested several lateral points on where we had the lowest deviation in finger widths, in order to fix a measuring point on our optical box.  Please look at the picture below for a better idea of how it works.

                                               

This is a picture of the placement of the pegs; the next picture is of where we decided to take our fix linear measuring.

 

               

We decided to take the measurements along a 14 cm slit, with a 0.5cm gap, and 2.5 cm distance from the index finger peg.  From this distance we could optimize our results.  Once we realized our strategically point we then decided to take more photocopies with the new measurements.  This would tell us the standard deviation of each subject’s hand’s position and the standard deviation of each finger’s width.  The charts show the results.

 

Person 2 with a deviation of 0.225 cm caused the highest position deviation.  From these results we calculated the distance the peg should be from the slot.  The next chart is the deviation of the six person’s fingers.

These results show that person number 6 had the greatest deviation in the heart finger, which was 0.0801 cm.  From these results we can now begin to notice that the position of the hand on the measuring slit and the widths of the fingers are related and need to be analyzed in order to achieve a good point of measure.  The measurements of the measuring slit where then chosen to have a 4.33% deviation on both position and finger width.

            There are basically three characteristics an ideal biometrical identification should have according to “The Biometric Group Inc”:

1)      Low Failure rejection

2)      Low false acceptance ratio

3)      Ergonomic and Ease of use

All of these characteristics have been accounted and maximized in our design.

Market Analysis:

            The biometrical market is not saturated and is quite open to new systems due to its intricate characteristics.  There are basically seven biometrical identification technologies in the market today. Below is a brief description of each technology, its applications, public acceptance and cost.

 

1)       Finger Scanning

TECHNOLOGY

Finger scan systems can be broadly categorized into two types: identification systems, known as AFIS (automatic fingerprint identification systems), and verification systems. Finger scan technology is based upon the fact that individuals' fingerprints have unique characteristics. These characteristics are whorls, arches, loops, ridge endings and ridge bifurcations.

Verification systems capture the flat image of a finger and perform one-to-one verification. The verification is performed in a few seconds.

PUBLIC ACCEPTANCE

Finger scan technology, more so than other biometrics, generates a lot of discussion regarding public acceptance issues. While it is fast and relatively easy to use, some people feel that by being required to submit a fingerprint, they are being treated like criminals. However, the case against finger scan technology is often overstated.

COST

Verification systems finger scan units range in cost from a hundred to a few thousand dollars, including hardware and software, depending upon the configuration. The advent of silicon technology, as well as the entry of such manufacturers as Sony, Motorola, and Infineon into the finger scan market, will drive prices down even further. 

2)       Face Geometry

TECHNOLOGY

A camera is used to acquire image of face from a distance of a few feet. The system then analyzes particular features or discreet areas of the face such as the distance between the eyes and the nose and the shape and location of the cheekbones. Most systems feature a face locating function that searches for faces within the field of view. Face recognition systems are designed to compensate for glasses, hats and beards.

The technology can perform verification and identification. Anecdotal evidence suggests that face recognition technology has the capability to be very accurate. However, the use of face recognition for one-to-many identification searches is relatively new and there is no reliable data to prove the accuracy rates.

PUBLIC ACCEPTANCE

Face recognition has a similar level of public acceptance as static photographs. It is possible to implement a biometric identification solution, which is transparent to the user. Users may be asked to stand for about 20 seconds to complete the enrollment process.

COST

The cost for a one-to-many face recognition application is similar to an AFIS finger scans application. The cost for a one-to-one verification system is approximately $100 per site for the software. The system will operate with a standard off-the-shelf video camera.

3)       Iris Recognition

TECHNOLOGY

Iris recognition technology involves the use of a camera to capture an image of the iris, the colored portion of the eye. The iris is an excellent choice for identification: it is stable throughout one’s life, it is not very susceptible to wear and injury, and it contains a pattern unique to the individual. Indeed, an individual’s right and left iris patterns are completely different.

There are two types of iris recognition systems: automatic capture and manual capture. In the manual system, the user must adjust the camera forward or backward a few inches in order to bring the iris into focus. Further the user must be within 6 – 12 inches of the camera. This requires substantial supervision and instruction. The automatic capture system incorporates a set of cameras to automatically locate the users face and eye, therefore removing the need to manually focus the camera. This system is substantially easier to use.

PUBLIC ACCEPTANCE

There are two key public acceptance issues: intrusiveness and ease of use. While the passive system is not physically intrusive, there are some people who are hesitant to use the system due to the perception that the camera is taking a picture of one’s eye.

Active iris scan requires more participation on the part of the user because the capture mechanism needs to be manually focused and the user must be close - approximately 3 inches - from the camera. The user sees a reflection or picture of the eye, and is thus more aware of what the system is doing. Although the technology is very similar to passive iris scan, the process is less transparent and thus can be seen as more intrusive.

COST

Iris recognition was traditionally among the most expensive biometric technologies, costing tens of thousands of dollars. The significant drop in the price of computer hardware and cameras, as well as the partnership between IriScan and LG, has brought the price of the high-end physical security unit into the $4000-$5000 range. 

The IriScan PC Iris, a proof-of-concept product showing that iris technology can be used be used in the home or office, is priced in the $700-$800 range. IriScan plans to release less expensive, easier-to-use products in the first half of 2000, and hopes to break the $500 price barrier.

4)       Signature Verification

TECHNOLOGY

In dynamic signature verification (DSV), the user signs his signature on a digitized graphics tablet. Signature dynamics, such as speed, relative speed, stroke order, stroke count and pressure are analyzed. The system compares not merely what the signature looks like, but also how the signature is signed. Dynamic signature verification is qualitatively different from electronic signature capture in which the signature is merely stored in electronic form and no biometric comparisons are performed.

DSV performs only one-to-one verification is considered one of the less accurate biometrics. The technology measures a behavioral characteristic and the user can easily change his signature to generate a false rejection.

PUBLIC ACCEPTANCE

DSV closely resembles the traditional signature process and has minimal public acceptance issues.

COST

The only hardware needed is the graphics tablet, which costs about $75 in quantity. The software cost would be determined based on the number of users.

5)      Voice Recognition

TECHNOLOGY

The user states a given pass phrase and the system creates a template based on numerous characteristics, including: cadence, pitch, tone, and shape of larynx. Speaker verification works with a microphone or with a regular telephone handset, although performance increases with higher quality capture devices. It is considered to be a hybrid behavioral and physiological biometric because although the voice pattern is determined to a large degree by the physical shape of the throat and larynx, the user can alter it. Background noise greatly affects how well the system operates.

The technology is considered to be far less accurate than fingerprint and iris scan technology. It is used solely for verification and requires user cooperation.

PUBLIC ACCEPTANCE

The technology is easy to use and does not require a great deal of user education. However, care must be taken to ensure that the user speaks at the appropriate time and that he speaks in a clear voice. The most important factor is to ensure that there is not too much background noise during enrollment or verification. Some users are self conscious about using the system because they feel that they are "performing".

COST

The cost of the system is based on the number of users. Usually speaker verification is an added function to an interactive voice response or call center application. The hardware costs are minimal as the technology works with regular telephones or PC microphones.

6)       Keystroke Dynamics

TECHNOLOGY

Keystroke Dynamics analyzes the characteristics of one's typing. It is a very new technology to the biometrics arena. Users enroll by typing the same word or word a number of times. Verification is based on the concept that the rhythm with which one types is distinctive. It is a behavioral verification system that works best for users who can "touch type". Currently NetNanny is working to commercialize this technology.

7)       Hand Geometry

TECHNOLOGY

There are three different technologies that look at the shape of the hands or fingers: hand geometry, single-finger geometry, and two-finger geometry. To use hand geometry the user places his hand on a platen and positions it by lining it up with 5 guide pegs. The system takes a picture of the hand and examines 90 characteristics, including three-dimensional shape of the hand, length and width of fingers and shape of knuckles.

For single-finger geometry, the user places his finger in a plunger and pushes forward into the device. The system has a set of rollers that roll around the finger and take measurements of 12 cross sections of a 1-½ inch span of the finger.

To use a two-finger geometry system, the user places the index and middle finger on a platen.

PUBLIC ACCEPTANCE

The finger/hand geometry systems do not raise many privacy issues and the technology is easy to use.

 

COST

The finger/hand geometry systems cost approximately $1500 per unit, depending on quantity and configuration.

APPLICATIONS

Hand geometry has been used for physical access and time & attendance at a wide variety of locations, including Citibank data centers, the 1996 Atlanta Olympics and New York University dorms. Lotus Development Corp. uses hand geometry to verify parents when picking up children from daycare. The University of Georgia uses hand geometry to verify students when they use their meal card.

One important application of government verification using hand geometry is INSpass. The Immigration and Naturalization Services department of the US government has rolled out an unmanned kiosk to expedite frequent travelers through customs. Enrolled users (limited to those considered ‘low-risk’) present their INSpass card and then submit their hand biometric sample. The system can currently be found in 8 airports, including San Francisco, New York, Newark, Toronto and Miami. The plan is to move to 20 airports by the end of 2000.

Single-finger geometry is currently designed only for physical access and time & attendance.

Two-finger geometry is designed with physical access with a major application set up for verification of seasons ticket holders at Disney World.

VENDORS

Recognition Systems, now a division of Ingersoll-Rand, has been selling hand recognition devices since the 1970s. MicroIdentification, a five year old company, manufactures single-finger geometry. BioMet Partners manufactures the two-finger geometry system.

 

                Below is a chart of the total revenue produced by each technology

 

“The biometric market is poised for a breakthrough to new levels of sales and visibility in the next 2-3 years, as four internal and external factors will combine to overcome what has been a long period of slower-than-expected growth. One of these factors is price reduction.” By IBG 2000 market report

This quote clearly states how biometrical identification systems are becoming more and more available due to their prices.  Our design meets this criterion by being a cheap identification solution. 

As the information above shows, there are only two hand geometry vendors in the US market.  This clearly shows that this market has been yet unexplored and we can penetrate this market with our biometrical device.

Below is an approximate cost of the parts we have used in our system.

Device

Cost

Microcontrollers

$60

CCD

$21

Resistors

$3

Capacitors

$3

Peripherals (Parallel port, Serial Port, cables, sockets, etc)

$6

Filtering IC’s

$2

Lamp

$2

Optical lens

$15

PCBoard

$8.95

Power Supplies

$20

Optical Box

$15

TOTAL

$155.95

Our total system cost is about $155.95.  The cost of all the system parts can be considerably reduced upon a revision to the system, since at first we bought several parts that we have chosen not to put into the design.

We now go into the description of the optical system this device contains.

 

 

OPTICAL SYSTEM

            The optical system of this device is responsible for the identification of an individual.  For this project, it is necessary to have a linear lamp that will project light parallel to the hand in measurement.  This lamp will produce dark shadows under the fingers to be measured.  External light will not have an effect on our system, since we are using a linear slit, and our CCD (Charged Coupled Device) will be inside a dark box.  This slit, as mentioned before in our statistical review, is 14cm in length and 0.5 cm in width.

 

Please look at the figure below:

 

                                       

This is a cut view of our system.  It shows how the slit is aligned with the angle of view of the lens.  The lens function is extremely crucial since it reduces the 14 cm gap to approximately 3.5 cm, which is the length of out CCD chip.  Now in order to correctly aligned the lens, an optical box was constructed.  The operation of the device is as follows: the pegs place the hand strategically and then the LED lamp is turned on. The image is then reduced by the lens and laid out on top of the CCD chip.  The CCD chip then transforms the light into voltage, approximately 4.5 volts, and the dark spots too, approximately 200mv.  This way the CCD only recognizes shadows and whites.  Exactly what we need to measure our finger’s widths.

            The Lens is extremely crucial and for this project we used a bi-convex lens, which reduced our image.  However, we had to recur to a lens manual in order to obtain the formula for the focal length of the lens.  This was 2 X the focal length + 1 cm.  From this measurement we constructed our box.

 

 

 

 

 

 

 


The diagram below is what happens when you place the hand on the slit and the lamp is turned on.

 

As you can see the lens, reduces the 14 cm image into a 3 cm image, but it inverses the image.  This is taken into consideration when we get the reading from the CCD and store it.

 

 

 

 

 

 

 

 

 

 

Electrical System Reviewed.

For this project we are using a CCD (Charged Couple Device) that has the following specifications:

Sony ILX703A

-          2048 pixel linear B/W CCD

-           Pixel Size ( 14 uM x 14 uM )

Resolution in our project:

Total pixel length = 2.8762 cm

Resolution per mm =  71 pixels / mm (Impressive!!)

 

Box Diagram of System:

 

 

 

 

How it works:

-          We have our host program running on the Computer.  When an administrator wants to do a scan of a person’s hand, he clicks on the appropriate button on the software and a signal is sent to the device’s microcontroller that is in standby mode.  When this signal is received, the microcontroller synchronizes the CCD and then waits for the user to press a button on the device indicating that the hand is ready to be scanned.  After the button is pressed, the CCD emits its Vout signals. These signals then enter a filter, where whites and shadows are detected, due to voltage differences; this is done through simple op-amps as peak detectors.  We then digitize the signal by passing it through and inverter.  This inverts the darks (Logical 0) to logical 1’s.  Then the microcontroller waits for the 0-1 transitions and counts the clock signals sent to the CCD. This determines the amount of pixels per finger.  These counts are stored into four different variables.  After the last 1-0 transition is detected by the microcontroller, we have then read the four widths of the fingers.  We now enter the writing phase.  In this phase we pass the values of each variable to the computer.  This is done through the parallel port, using 4 data lines. Now each finger will have a length of 9 bits.  2048 / 4 = 512, therefore 9 bits.  So we would need to pass 4 bits at a time to the computer, so 3 cycles will be needed per finger.  In total, it would be 12 cycles per scan.

 

Micro-Controller Software Review:

 

            There are basically 5 states in the programming of the microcontroller.

 

1.      Standby mode

2.      Scan mode

3.      Synchronization mode

4.      Reading mode

5.      Write mode

 

1)      Standby mode: This mode consists on a continuous loop turning a green LED on the device, to indicate activity, and constant monitoring of the parallel port for a SCAN signal.

2)      Scan Mode: We turn our LED on, to indicate reading, and we turn on the lamp.  We then wait for approximately 200ms in order for the lamp to reach full brightness. Then we go to synchronization mode

3)      Synchronization mode: In this mode we send the appropriate clock signals to the CCD, in order to obtain an output.  Look at attached time diagram.  We then create a counter for the clock signal and we count 33 dummy pixels.  We then enter reading mode.

4)      Reading mode: We then wait for a 1 to 0 transition of Vout, indicating a shadow (finger1) and start counting the clock cycles.  When we reach a 0 to 1 transition again, we have finished reading finger 1.  We store the counter’s contents into variable X1 and repeat the process until we have counted the four finger’s width.  We then enter the Write mode.

5)      Write Mode: First, we turn off the LED to indicate finalization of the scanning mode.  We then start sending our variables through the parallel port to our host software on the PC.  We then turn our lamp off and sound our buzzer for 2 seconds, to indicate the end of the scanning period.

 

PC Software Review:

 

            Our host program was coded in Visual Basic and our database was done in Microsoft Access.  The software contains several options:

1)      Welcome Screen offering help

2)      Add/Remove Users from the database

3)      Scan

4)      Options

 

Here are some Screenshots of the program:

 

 

 

 

1)      Our welcome screen contains detailed information on how to use the software and the device.  It contains some help screens and troubleshooting guides.

2)      Add/Remove User Panel:  From this panel you can add and remove users from the database.  When you press the add user button, you get a form that you fill out with the user’s information.  Also, you add a PIN number for each user, which is unique.  This PIN number is the key to identifying a user.  You then go to a Scan form, where the user enters his/her pin, then places his hand on the device, and will get 10 readings of his/hers hand.  We decided on 10 trials, in order to get enough information for our detection algorithm to successfully recognize the user.  To remove a user, you get a list of all the users in the system and then click on the desired user to remove, then press the remove button and reconfirm for deletion.

3)      Scan Option:  This tab is where user recognition takes place.  First, a user input his/her Pin number.  Then he is told to place his/her hand on the scanner and press the scan button on the device when ready.  A scan is produced and then compared to the databases information for approval or rejection.

4)      Options Tab:  In the options tab, the identification administrator can choose the type of detection algorithm to choose for the whole system.  There are two choices:

a.       Static Detection

b.      Dynamic Detection

 

 

Algorithm Explanation

            This section contains the explanation for the detection algorithms used in our system.

1)      Static Detection:  This is the easiest, but less reliable for intruder rejection, algorithm. It basically consists on the following:  Each user is first scanned 10 times in order to have an active database of measurements active.  After we have our ten measurements we take all of these ten results and obtain a Medium for each finger.  This is the average of all the ten results for each finger.  We then set a standard error of 20% on each finger.  That is why it is static; you can choose the deviation percent.  When a user then scans for an approval or rejection the following takes place:  We obtain the first fingers measurements and compare them to the medium, using a delta+ of 20% and a delta- of 20%.

 

    

 

If the finger result is within bounds, then the first finger is accepted.  We then do is for the remaining three fingers with their appropriate medium and the 20% deviation.

 

One can observe that some people may have more deviations to the right of the medium or to the left.  This depends on their finger’s build.  That is why the dynamic algorithm is created.

 

 

2)      Dynamic Detection:  This is a more complex algorithm since we now use more of the user’s finger dimensions to obtain rejection or acceptance results. Look at diagram below:

 

 

 

This diagram shows how a user has a bigger percentage on the delta- than on the delta+ variable.  This is the key to the algorithm!!!  However, in order to produce effective results, we must have at least 50 readings for a single user.  Our software does this automatically, since we are always keeping the 100 relevant results for each finger.  In a manner of speaking our software learns by itself.  For this algorithm, first a medium is calculated with, e.g. 50 results.  We then obtain the 8 biggest reading for a finger and discard them, and vice versa for the 8 smallest values.  We then obtain an average of the remaining highest results and obtain a second medium, called Delta +.  We do the same for the small values and obtain a delta -.  Now we have a dynamic scale of measurement for comparison.  This is dynamic thresholding.  With this algorithm we can now reduce our rejection error and increase our false acceptance ratio.  Since each person will have a different scale of comparison. 

The user can choose from the options menu, the percentage to discard for each delta value.  This concludes our dynamic algorithm.

 

 

In conclusion, we have seen how our linear biometric identification system works.  It is a linear hand geometry reader, which possesses the same basic characteristics of any 3-D hand geometry system.  Although the system relies on the CCD resolution, it is almost imperative to have a good detection algorithm in play.  That is why we have employed two detection algorithms: static and dynamic that have been explained above.  These algorithms are the sole key to producing and effective and reliable identification system.  These algorithms are based on the statistical properties of every user’s fingers and are therefore unique for each person.  However, this system is no complete.  We have written down several optimization criterions that we could add to this versatile system.

1)      Stand-alone System:  The integration of a data ROM, to include the user database and a System BIOS to contain the Operating System of the device (synchronization procedures, scanning, writing, etc.)

2)      An upgrade to the optical lens: In order to reduce the size of the optical box.

3)      The use of a laser light source in order to have no noise on the CCD due to light.  This way a light filter will not be necessary.

4)      The incorporation of a keypad for system input.  This will include PIN entry.

5)      An LCD display to let the user know the results of their scans.

6)      A bar code reader.  This will implement the use of a PIN card.  Which includes the user’s basic info and will locate the user’s entry on the database upon reading.

These are some optimizations that can be done to this system in order to provide a more reliable and secure identification system.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BIBLIOGRAPHY

 

·         A. K. Jain, L. Hong and S. Pankanti, "Biometric Identification", Comm. ACM, pp. 91-98, Feb. 2000.

 

·         K. Jain, A. Ross and S. Prabhakar, " Biometrics-Based Web Access", MSU Technical Report TR98-33, 1998.

 

·         Biometric Market Group “Biometric Market Report 2000”, BMG January 2000

 

·         International Biometric Group “Biometrics: An overview” December 1999

 

·         Sandia National Laboratories “The Sandia Report on Biometrical Identification” 1991