CA2050893A1 - Finger profile identification system - Google Patents
Finger profile identification systemInfo
- Publication number
- CA2050893A1 CA2050893A1 CA002050893A CA2050893A CA2050893A1 CA 2050893 A1 CA2050893 A1 CA 2050893A1 CA 002050893 A CA002050893 A CA 002050893A CA 2050893 A CA2050893 A CA 2050893A CA 2050893 A1 CA2050893 A1 CA 2050893A1
- Authority
- CA
- Canada
- Prior art keywords
- finger
- profile
- candidate
- edges
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Classifications
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C9/00—Individual registration on entry or exit
- G07C9/20—Individual registration on entry or exit involving the use of a pass
- G07C9/22—Individual registration on entry or exit involving the use of a pass in combination with an identity check of the pass holder
- G07C9/25—Individual registration on entry or exit involving the use of a pass in combination with an identity check of the pass holder using biometric data, e.g. fingerprints, iris scans or voice recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/107—Static hand or arm
- G06V40/11—Hand-related biometrics; Hand pose recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
Abstract
A method of and apparatus for verifying the identity of an individual is based on the profile of a hand print. A set of reference data is created by viewing (9, 10) an individual's hand and determining the location of the edges of each finger separately to generate a set of finger profiles, establishing a central axis for each finger profile by reference to the edges thereof, rotating the central axis of each finger profile until it is normal to a reference line, determining the width of each finger profile at predetermined locations parallel to said reference line, and recording said finger profile information in a memory such as an ID card.
Thereafter, a candidate's finger profile can then be compared (6) with the reference profile obtained by reader/recorder (7) by comparing the width of each finger profile at predetermined locations, determining whether the correspondence between finger widths at said predetermined locations meets a predetermined criteria, and then indicating acceptance or rejection of the candidate.
Thereafter, a candidate's finger profile can then be compared (6) with the reference profile obtained by reader/recorder (7) by comparing the width of each finger profile at predetermined locations, determining whether the correspondence between finger widths at said predetermined locations meets a predetermined criteria, and then indicating acceptance or rejection of the candidate.
Description
W O 90/t2371 PC~r/US90/01982 - 1- 2~8~3 FINGER PROFILE ID~NTIFICATION SYSTEM
B~ckGroUnd of the Invention A number of devices currently exist which employ biometrics to verify an individual`s identity for s~curity purposes. While some of these devices have been marketed for commercial purposes, none has yet to be found to be useful, reliable, and~or economical enough for large-scale adoptlon.
Summary of the Invention This invention relates to novel method and apparatus for verifying the true identity of an individual seeking entry to a secured installation. The method uses the statistically valid theory that an individual's hand geometry, such as finger size and shape, are different ~rom person-to-person, much like fingerprints. ~n individual's finger profiles are recorded in a memory that may be either on a credit-card-type ID or elsewhere. When that person subsequently applies for entry to a secured installation requiring the ID, the identity is verified by a comparison between the current finger profiles and the reference profiles in the memory. The system then recommends an accept or reject decision based on one or more statistical crite`ria.
B~ckGroUnd of the Invention A number of devices currently exist which employ biometrics to verify an individual`s identity for s~curity purposes. While some of these devices have been marketed for commercial purposes, none has yet to be found to be useful, reliable, and~or economical enough for large-scale adoptlon.
Summary of the Invention This invention relates to novel method and apparatus for verifying the true identity of an individual seeking entry to a secured installation. The method uses the statistically valid theory that an individual's hand geometry, such as finger size and shape, are different ~rom person-to-person, much like fingerprints. ~n individual's finger profiles are recorded in a memory that may be either on a credit-card-type ID or elsewhere. When that person subsequently applies for entry to a secured installation requiring the ID, the identity is verified by a comparison between the current finger profiles and the reference profiles in the memory. The system then recommends an accept or reject decision based on one or more statistical crite`ria.
2~ 3 - 2 -Yet another fea~ure of the invention is that the same device may be used with x-ray images of the fingers, or other body parts ~instead of video images o live fingers), wherein bone shapes derived from previously stored x-ray ~5 images are compared with those of recent images. This ability may be useful for forensic purposes such as identifying combat~casualties or accident victims.
Anticipated uses for the sy~tem include milltary bases, intelligence facilities, nuclear installations, bank vaults, computer rooms and computers; preventing use of automatic bank teller machines by means of lost, stolen, or counterfeit I.D. cards; and preventing timecard fraud.
Both the mechanism and software or firmware required for performing the above computations to accomplish an entry decision are remarkably simple and inexpensive.
A feature of the invention is the ability of the system to accommodate various hand placements. Also, the system employs an autocorrelation procedure which further accommodates normal age-related variations in the finger profile. As long as an individual does not have a degenerative joint disease, the initial finger shape data can serve as his ID for many years without updating.
A method of verifying the identity of an individual based on the profile of a hand print, the method comprising the steps of creating a set of reference data by 8 ~ 3 a) determining the location of the edges of each finger separately thereby to generate a set of finger profiles, b) establishing a central axis for each finger profile by reference to the edges thereof, c) rotating the central axis of each finger profile until it is normal to a reference llne, d~ determining the width of each inger profile at predetermined locations parallel to said reference line, and e) recording said finger profile information; and thereafter determining a candidate's finger profile in accordance with steps a) to d) above, comparing the candidate's finger profile with the reference profile by comparing the width of each finger profile at predetermined locations, determining whether the correspondence between finger widths at said predetermined locations meets a predetermined criteria, and indicating acceptance or rejection of the candidate. It is also an object of this invention to provide an apparatus for performing these method steps.
Other objects and advantages of the invention will be apparent from the following description, the accompanying drawings and the appended claims.
~ ings Fig. l is a block diagram showing the components comprising the invention.
2~50893 4 _ Fig. 2 is a cross-sectional side view of a flnger scanning device constructed according to the invention.
Fig. 3a shows a hand profile of four fingers as typically placed on the back lighted panel with the scanning lines passing across each finger.
Fig. 3b shows a hand profile after the fingers have been mathematically rotated.
Fig. 4 is a block diagram of the basic components making up a computer processor.
Figs 5a. 5b and 5c are curves representing the autocorrelation process where Fig. 5a shows the front slither, Fig. 5b shows the middle slither and Fig. 5c shows the back slither.
Descri~tion of the Pr~f~rL~ Embodiment Referring now to the drawings which depict a biometric personnel verification system for screening entry access to controlled areas or equipment, the device shown in Figs~ 1 and 2 includes a video camera 10 for generating a high-resolution silhouette of an individual's finger shapes and sizes. This silhouette is converted into a numerical array of width values for each finger, taken at predetermined intervals lsuch as 1 mm) from interdigital web to fingertip. A typical profile of a single finger might WO~Otl2371 PCTJUS90/01982 ~ 5 ~ 3 consist of 50 to lOO width values. Profiles of four to eight fin~ers are recorded; thumbs are generally not used.
Upon enrollment of an authorized individual, the device stores finger shape and size reference data derived S from high-resolution video scan imagery on a RAN-type encoded ID card or in a secure data base. Thereafter, when seeking a~ceptance to a secure area. an individual inserts his ID card in the card reader 7, enters on keyboard 13 a personal identification number (if required), and places his hand on a panel or stage 9 where it is viewed by the solid-state video camera lO. The device then scans the individual's fingers and generates finger shape candidate data.
The previously stored finger profile information, l~ or reference data, is then compared with the current finger profile, or candidate data. To compare profiles, the system uses characteristics referred to herein as discriminants, which are algorithms that measure sameness. The numerical score resulting from a comparison ranges in value from 0.000 to l.OOOt a perfect match would be l.OOO. More than one discriminant may be used to create the match score to increase the reliability of the process.
When initially applying for an ID card, an authorized person inserts a blank ID card into slot 12 of card reader/recorder 7 and places his hand, fingers spread, WO90/1~71 PCT/US90/01982 2~ 3 - 6 -on the stage 9, a backlighted translucent panel.
Ba~klighting is provided by an incandescent, luminiscent, or fluorescent light source 1.
The image o~ the ~ingers is reflected off mirror 2 and through lens 3 onto motor-driven mirror 4 which sweeps the image in l-mm steps across CCD linear sensor array 5.
The ~mage data output of 5 is digitized by flash A/D 15, Fig, 4, applied to a computer processor 6, line-by-line where the individual finger edges are separated into left and right paired arrays and the center lines of all fingers are determined.
The silhouette shown in Fig. 3a is in the form of left and right paired coordinates at l-mm intervals comprises the outlines of four fingers, excluding the thumb.
These eight arrays are stored temporarily in computer memory 6, then`rotated as illustrated in Fig. 3 to correct for angular deviations owing to finger-spread, and then stored in computer memory 6. The process for finger rotation or normalization will be described below.
Each stored array is then read from memory in 1-mm slices from interdigital web to fingertip, yielding a sequence of width values which are perpendicular to the central axis of the finger (Fig. 3b).
The set of about 75 to 100 width values for each finger, called the profile, is sent to the card 2 ~
reader/recorder 7. The shape information called re~erence profiles is now stored in the card's random access memory lRAM) or magnetic or optically encoded strip on the back of an ID card which was insexted into the card slot 12 o the card recorder/reader 7. Alternatively, reference profile data may be stored in ~ secure data base.
A secure facility has a similar device ~s described above. When applying for entry to the facility, the candidate inserts the ID card into the slot of 7 which now acts as a card reader, enters an assigned personal identification number (PIN) via keyboard 13, Fig. l, and places his hand on backlighted panel 9. Solid-state video camera lO reads each finger silhouette and generates the candidate profiles of the fingers as described above.
lS A computer processor 6, Fig, 4, then compares the pre-stored reference profiles originally obtained from card reader 7 with the current candidate profiles by means of a battery of discriminants. The shape and size discriminants are chosen to be as nearly independent as possible f rom one another so that the quality of the match may be tested using a number of distinct parameters. A composite of the discriminant scores from all the fingers, called the Grand Match Number, ranges in value from 0.000 to l.000 with a perfect match being l.000.
.
Prior trials determine the score which must be achieved to declare acceptance of the candidate. A display panel indicator 14, Fig. 4 announces whether the candidate i8 accepted or rejected. (The display panel, may also provide other messages, such as advising the candidate to reposition the hand.) Subsequent to the accept/reject decision, an activator mechanism 11 such as a solenoid may operate an alarm, a door latch, turnstile, or other entry device.
Since the finger placement during enrollment and at the time of reguested entry will not be the same, a direct comparison between the data would not be meaninyful.
Therefore, this invention provides a method for rotating and translating the data obtained from the finger scans, that is figuratively to rotate the fingers so that they are all essentially parallel and that the slices recorded and compared are normal to the central axis of each finger, both at enrollment and later on.
Normalization of ~and Data Nhen the camera scans a hand, the data produced is filtered by hardware and intelligent software. The result is a list, for each scanline, of the places where the image brightness goes from dark to light and vice versa. The list is a set of numbers whose values are the pixel numbers at ~VO90~t2371 PCT/US90/01982 2 ~ 9 3 g which these transitions occur. This list is conditioned so that certain assumptions may always be made concerning the number of transitions on each scanline.
After the transition list is obtained. the program separates the list into fingers, i.e. each point in the list is ident~fied with the right or left side of a particular finger. The data for each inger is then normalized. This process involves rotating and translating the finger data points in the two dimensional plane (B2) so that the central axis of the finger is identified with the vertical axis and the pair of points (right and left side) on the finger that have the lowest value in their vertical coordinate (after the rotation has been performed) lie on the horizontal axis.
Finger Separation The transition list must be sorted by finger, and each finger sorted by right and left side. This accomplished by following the perimeter of the hand and looking at the extreme points (with respect to the vertical axis) of the image. The data has been previously conditioned so that it may be assumed that:
1) Each scanline has no more than 8 transitions.
2) Each scanline has an even number of transitions.
WOso/12371 PCT/US~0/01982 ;, '~ U ~ 3 3 - lo -3) There are no radical discontinuities in the image.
With this in mind, it is assumed that the first transition on the first scanline belongs to finger 1. The 5 algorithm then proceeds around the perimeter of the hand lmage. The tip and web of the fingers are identified with local maxima and minima, respectively~
Finger Rotation After the transition list has been reorganized into fingers, each finger is normalized. This is accomplished by performing a number of operations on the transition list data for the finger, these are:
1) Change the coordinate system of the data from pixels to hundredths of a mm.
2) Find the best center line through the finger.
Since the transition list for the finger has already been organized into right and left sides, the location of the first (proceeding from the base to the tip of the finger) actual right/left pair must be found. After this is accomplished, one may assume (because of the previous data conditioning) that the rest of the data points are paired.
~t is simple to find the coordinates of the midpoint between each pair. A list of these midpoints is made and passed to a modified linear regression (one which minimizes the sum of WO90/123~1 PCT/US90/01982 tbe squares of the perpendicular distance from each point to the best line. as opposed to the sum of the squares of the vertical distance from each point to the best line).
3) Use the slope of the center line, m, eo calculate an orthonormal basis for R2, B' ~ xl.x2t, where xl i8 a unit vector that points along the center line of the ~ng~r from base to tip, and x2 is a unit vector perpendicular to xl s~ch that the crossproduct of xl and x2 yields a vector in the positive Z direction (right hand coordinate system).
Anticipated uses for the sy~tem include milltary bases, intelligence facilities, nuclear installations, bank vaults, computer rooms and computers; preventing use of automatic bank teller machines by means of lost, stolen, or counterfeit I.D. cards; and preventing timecard fraud.
Both the mechanism and software or firmware required for performing the above computations to accomplish an entry decision are remarkably simple and inexpensive.
A feature of the invention is the ability of the system to accommodate various hand placements. Also, the system employs an autocorrelation procedure which further accommodates normal age-related variations in the finger profile. As long as an individual does not have a degenerative joint disease, the initial finger shape data can serve as his ID for many years without updating.
A method of verifying the identity of an individual based on the profile of a hand print, the method comprising the steps of creating a set of reference data by 8 ~ 3 a) determining the location of the edges of each finger separately thereby to generate a set of finger profiles, b) establishing a central axis for each finger profile by reference to the edges thereof, c) rotating the central axis of each finger profile until it is normal to a reference llne, d~ determining the width of each inger profile at predetermined locations parallel to said reference line, and e) recording said finger profile information; and thereafter determining a candidate's finger profile in accordance with steps a) to d) above, comparing the candidate's finger profile with the reference profile by comparing the width of each finger profile at predetermined locations, determining whether the correspondence between finger widths at said predetermined locations meets a predetermined criteria, and indicating acceptance or rejection of the candidate. It is also an object of this invention to provide an apparatus for performing these method steps.
Other objects and advantages of the invention will be apparent from the following description, the accompanying drawings and the appended claims.
~ ings Fig. l is a block diagram showing the components comprising the invention.
2~50893 4 _ Fig. 2 is a cross-sectional side view of a flnger scanning device constructed according to the invention.
Fig. 3a shows a hand profile of four fingers as typically placed on the back lighted panel with the scanning lines passing across each finger.
Fig. 3b shows a hand profile after the fingers have been mathematically rotated.
Fig. 4 is a block diagram of the basic components making up a computer processor.
Figs 5a. 5b and 5c are curves representing the autocorrelation process where Fig. 5a shows the front slither, Fig. 5b shows the middle slither and Fig. 5c shows the back slither.
Descri~tion of the Pr~f~rL~ Embodiment Referring now to the drawings which depict a biometric personnel verification system for screening entry access to controlled areas or equipment, the device shown in Figs~ 1 and 2 includes a video camera 10 for generating a high-resolution silhouette of an individual's finger shapes and sizes. This silhouette is converted into a numerical array of width values for each finger, taken at predetermined intervals lsuch as 1 mm) from interdigital web to fingertip. A typical profile of a single finger might WO~Otl2371 PCTJUS90/01982 ~ 5 ~ 3 consist of 50 to lOO width values. Profiles of four to eight fin~ers are recorded; thumbs are generally not used.
Upon enrollment of an authorized individual, the device stores finger shape and size reference data derived S from high-resolution video scan imagery on a RAN-type encoded ID card or in a secure data base. Thereafter, when seeking a~ceptance to a secure area. an individual inserts his ID card in the card reader 7, enters on keyboard 13 a personal identification number (if required), and places his hand on a panel or stage 9 where it is viewed by the solid-state video camera lO. The device then scans the individual's fingers and generates finger shape candidate data.
The previously stored finger profile information, l~ or reference data, is then compared with the current finger profile, or candidate data. To compare profiles, the system uses characteristics referred to herein as discriminants, which are algorithms that measure sameness. The numerical score resulting from a comparison ranges in value from 0.000 to l.OOOt a perfect match would be l.OOO. More than one discriminant may be used to create the match score to increase the reliability of the process.
When initially applying for an ID card, an authorized person inserts a blank ID card into slot 12 of card reader/recorder 7 and places his hand, fingers spread, WO90/1~71 PCT/US90/01982 2~ 3 - 6 -on the stage 9, a backlighted translucent panel.
Ba~klighting is provided by an incandescent, luminiscent, or fluorescent light source 1.
The image o~ the ~ingers is reflected off mirror 2 and through lens 3 onto motor-driven mirror 4 which sweeps the image in l-mm steps across CCD linear sensor array 5.
The ~mage data output of 5 is digitized by flash A/D 15, Fig, 4, applied to a computer processor 6, line-by-line where the individual finger edges are separated into left and right paired arrays and the center lines of all fingers are determined.
The silhouette shown in Fig. 3a is in the form of left and right paired coordinates at l-mm intervals comprises the outlines of four fingers, excluding the thumb.
These eight arrays are stored temporarily in computer memory 6, then`rotated as illustrated in Fig. 3 to correct for angular deviations owing to finger-spread, and then stored in computer memory 6. The process for finger rotation or normalization will be described below.
Each stored array is then read from memory in 1-mm slices from interdigital web to fingertip, yielding a sequence of width values which are perpendicular to the central axis of the finger (Fig. 3b).
The set of about 75 to 100 width values for each finger, called the profile, is sent to the card 2 ~
reader/recorder 7. The shape information called re~erence profiles is now stored in the card's random access memory lRAM) or magnetic or optically encoded strip on the back of an ID card which was insexted into the card slot 12 o the card recorder/reader 7. Alternatively, reference profile data may be stored in ~ secure data base.
A secure facility has a similar device ~s described above. When applying for entry to the facility, the candidate inserts the ID card into the slot of 7 which now acts as a card reader, enters an assigned personal identification number (PIN) via keyboard 13, Fig. l, and places his hand on backlighted panel 9. Solid-state video camera lO reads each finger silhouette and generates the candidate profiles of the fingers as described above.
lS A computer processor 6, Fig, 4, then compares the pre-stored reference profiles originally obtained from card reader 7 with the current candidate profiles by means of a battery of discriminants. The shape and size discriminants are chosen to be as nearly independent as possible f rom one another so that the quality of the match may be tested using a number of distinct parameters. A composite of the discriminant scores from all the fingers, called the Grand Match Number, ranges in value from 0.000 to l.000 with a perfect match being l.000.
.
Prior trials determine the score which must be achieved to declare acceptance of the candidate. A display panel indicator 14, Fig. 4 announces whether the candidate i8 accepted or rejected. (The display panel, may also provide other messages, such as advising the candidate to reposition the hand.) Subsequent to the accept/reject decision, an activator mechanism 11 such as a solenoid may operate an alarm, a door latch, turnstile, or other entry device.
Since the finger placement during enrollment and at the time of reguested entry will not be the same, a direct comparison between the data would not be meaninyful.
Therefore, this invention provides a method for rotating and translating the data obtained from the finger scans, that is figuratively to rotate the fingers so that they are all essentially parallel and that the slices recorded and compared are normal to the central axis of each finger, both at enrollment and later on.
Normalization of ~and Data Nhen the camera scans a hand, the data produced is filtered by hardware and intelligent software. The result is a list, for each scanline, of the places where the image brightness goes from dark to light and vice versa. The list is a set of numbers whose values are the pixel numbers at ~VO90~t2371 PCT/US90/01982 2 ~ 9 3 g which these transitions occur. This list is conditioned so that certain assumptions may always be made concerning the number of transitions on each scanline.
After the transition list is obtained. the program separates the list into fingers, i.e. each point in the list is ident~fied with the right or left side of a particular finger. The data for each inger is then normalized. This process involves rotating and translating the finger data points in the two dimensional plane (B2) so that the central axis of the finger is identified with the vertical axis and the pair of points (right and left side) on the finger that have the lowest value in their vertical coordinate (after the rotation has been performed) lie on the horizontal axis.
Finger Separation The transition list must be sorted by finger, and each finger sorted by right and left side. This accomplished by following the perimeter of the hand and looking at the extreme points (with respect to the vertical axis) of the image. The data has been previously conditioned so that it may be assumed that:
1) Each scanline has no more than 8 transitions.
2) Each scanline has an even number of transitions.
WOso/12371 PCT/US~0/01982 ;, '~ U ~ 3 3 - lo -3) There are no radical discontinuities in the image.
With this in mind, it is assumed that the first transition on the first scanline belongs to finger 1. The 5 algorithm then proceeds around the perimeter of the hand lmage. The tip and web of the fingers are identified with local maxima and minima, respectively~
Finger Rotation After the transition list has been reorganized into fingers, each finger is normalized. This is accomplished by performing a number of operations on the transition list data for the finger, these are:
1) Change the coordinate system of the data from pixels to hundredths of a mm.
2) Find the best center line through the finger.
Since the transition list for the finger has already been organized into right and left sides, the location of the first (proceeding from the base to the tip of the finger) actual right/left pair must be found. After this is accomplished, one may assume (because of the previous data conditioning) that the rest of the data points are paired.
~t is simple to find the coordinates of the midpoint between each pair. A list of these midpoints is made and passed to a modified linear regression (one which minimizes the sum of WO90/123~1 PCT/US90/01982 tbe squares of the perpendicular distance from each point to the best line. as opposed to the sum of the squares of the vertical distance from each point to the best line).
3) Use the slope of the center line, m, eo calculate an orthonormal basis for R2, B' ~ xl.x2t, where xl i8 a unit vector that points along the center line of the ~ng~r from base to tip, and x2 is a unit vector perpendicular to xl s~ch that the crossproduct of xl and x2 yields a vector in the positive Z direction (right hand coordinate system).
4) Build a linear transformation, T that maps the Basis B' to the ~Standard) Basis, B - el,e2t ~ (1,0), (O,l)t. The transformation T is a rigid motion of the plane, i.e. the image will not be distorted by the transformation.
5) Apply T to the finger data. This will rotate the finger so that the finger center line is identified with the vertical axis and the vector pointing from the base to the tip of the finger will point. in the positive direction.
6) Perform a vertical axis shift on the data points so that the lowest pair of right/left points (those with the smallest vertical coordinate) on the finger lie on the horizontal axis.
7) Change vertical coordinates to mm.
` ` "`2 ~
These seven steps can be comblned into onQ simple operation. First, let ~ be the linear transformat~on on ~2 that rotates the flnger around the origln (0,0), 80 that the Qenter line of the finger is superimposQd on the vertlcal axis. In addition to the rotation, the flnger mu~t al80 be trsnslatQd to a standardizQd position rel~tive to the orlgln. ~RQmembQ~ that it was rotated around the origin, l~aving it in tbe proper orientat~on, but at some indeterminate posit~on on the plane~) To ~olve thls problem, T ~8 used to build another type of function. called an affine transformation. An affine transformation performs both rotation~ and tran~lations on its domain. The function S can be represented by a 2x2 matrix and the affine transformation u~ed can be represented by a 3x3 matrix a~-rll rl2 o t~r~ ~ r21 r22 o (t~r] could be thou~ht of as a 0 0 1 rotation matrix) w~erQ, ~T~ ~ ¦ rll rl2 ¦
I r21 r22 1 .
The above affine transformation performs only the transformAtion T, and not a translation. The mAtrix of an affine transformatlon to perform a s~mple translatlon would look l~ke:
1 o tx o o 2 ~
where, tx - the trans~ation performed along the x axis.
ty = the translation along the y axis.
The matrix of an affine transformation that will change coordinate systems would look like:
dx 0 0 ~c~ ' 0 dy 0 O O 1 ~
It is a nice property of matrices land of linear ` lO and affine transformations) that instead of performing steps l through 7 one at a time, one can combine many of these steps into one function (matrix). For this problem it is effectivè to find the center line first (before changing the coordinates of all the data points to hundredths of mm) and then build the rotation transformation A whose matrix is described above. Using this and other matrices, perform the following operations ~order is significant; matrices do not in general commute):
l) Change coordinate systems to hundredths of mm.
2) Translate the center of mass of the finger to the origin.
3) Rotate finger around origin (now center of mass).
4) Translate finger along the vertical axis so 2S that the lowest left-right pair lies on the horizontal axis.
2 ~
The matrix representation of this process would look like:
tF~] ~ 2][~,][At~A~ F~
Where the matrices of the affine transformatlons look like:
¦ dxl o 0 1 ¦ l 0 txl I
O O 1 I [A~O 1 tyl I
rll rl2 0 1 1 l 0 0 ~,] ' r21 r22 0 ~t2] - O 1 tY2 where, dxl = horizontal dimension of a pixel in hundredths of mm.
dyl = vertical dimension of a pixel in hundredths of mm.
txl 5 -(center of the fingers mass along horizontal axis).
tyl - -(center of the fingers mass along vertical axis).
rll = r22 = m/mag_m.
rl2~ = -r21 = -l/mag m.
m = slope of finqer center line.
mag m - +/- sqrt(m ~ m ~ l) Note: If m < 0 then make mag m < 0 tY2 ~ distance lin hundredths of mm) to shift the finger along the vertical axis tas in 4 above).
The data points of the finger F are represented in the matrix:
1 fxl fx2 fx3 ... fxn ¦
[F~ = fy1 fY2 fY3 fYn 1 1 1 ... 1 2 ~ 3 and each column vector (fxl,~y~ is a point on ~he S finger P.
The eguation:
lFro~m] ~ tAt2~ tA~] [Atl] tAcl] tF] -lO can be reduced to a single matrix multiplication, lF~ ][F]-where [A ] - [At2][A~][Atl][A&l].
It is desirable to calculate lA*] before working on F, for doing so greatly reduces the number of calculat~ons necessary for the computation.
After applyi*g A* to F, change the vertical coordinates from hundredths of mm to integer format mm.
Tbis obviously involves a loss of precision. To diminish the impact of this look at each mm section and average the horizontal coordinates of the points (left and right side separately, of course) that lie within the section to get the horizontal coordinate of the point at that section.
WO90/t237t PCT/US90/01982 ~ - 16 -2`~
After this reduction in resolution, data is checked for discontinuities.
The following items constitute a description of various discriminants and their derivations.
Discriminant I - Width The simplest discriminant is the coefficient of determination (r2) wherein corresponding slice widths between the reference and the candidate profiles are compared. Because the candidate's fingers may not be placed in the same positions on the panel as the reference fingers l~ a means must be followed to bring the two profiles into register so that they can be compared. This is done by shifting the profiles with respect to each other one slice at a time while testing the value of r2 at each shift until it reaches a maximum. At perfect match between identical profiles, the value of r2 ~ l.000. The point of best alignment between any two profiles is referred to as register. The numerical value of r2 at register is a meaæure of sameness of the profiles and is called Profile r2 W~90/t2371 PCT/US90/01982 - 17 _ 2~ 3 Discriminant II - Shape Prominences The first step is to locate and record tha shape prominences (joints and mid-shaft depressions) of the fingers as revealed by the reference profiles. The technique to do this is called autocorrelation. The second step is to do the same for the candidate profiles. ~he third step is to compare the locations of the correspond~ng prominences between candidate and reference autocorrelations. To extract these salient shape features from a profile, the autocorrelation procedure compares an original parent profile with itself in the form of a clone.
This has the effect of exaggerating the subtle prominences at the joints and depressions at the midshafts of the fingers.
An advantage of the autocorrelation procedure is that it forgives or accommodates any changes in the fingernail length, gross finger shape due to edema, or hand pressure, as long as such changes do not obliterate joint features altogether. Prominences located by autocorrelation àlso permit long-term identification despite normal variation in finger shapes due to aging. ~owever, it is recognized that in some persons with severe. progressive osteodegenerative disease such as arthritis, the reference profiles may have to be updated periodically.
WOgO/123~1 PCT/US90/01982 2 a~ aS~3 - 18 -The following is a description of the autocorrelation procedure which is carried out with the reference profiles and with the candidate profiles to be compared. The basic idea is described irst and a modified form is described thereafter.
A clone is made of an original (parent) profile, and an autocorrelation match curve is generated by measuring successive values of r2 as the clone profile is shifted stepwise relative to the parent profile by a distance at least as great as the separation between distal ànd proximal joints.
When the clone is in exact register with the parent profile, a peak is noted on the autocorrelation graph where the value of r2 ~ 1~000. Register, then, becomes a reference point to which prominent features of the autocorrelation curve can be referred.
Shifting the clone in l-mm intervals along the length of the parent profile, causes the value of r2 to decline as the two profiles become progressively further out of register. Minimum r2 (nadir) occurs when the distal prominence of the clone aligns with the midshaft depression of the parent. The number of l-mm intervals from register re~uired to reach thiQ nadir, determines the distance between the distal interphalangeal joint and the midshaft of 2~a~
the middle phalanx, and this value (in mm) is recorded for future reference.
Shifting the clone past the nadir to the point where the distal joint prominence of the clone lines up with the middle ~nterphalangeal p~ominence o~ the parent, locates a pseudo-match (zenith), a local maximum in the value of r2.
The number of l-mm steps from register to this zenith i8 the distance between the distal and middle interphalangeal joints. Further shifting ~misalignment) may locate yet another nadir relative to register. The number of matchable points (n) declines with each shift away from register, and this undesirable condition introduces some artifacts in the autocorrelation curve.
So, to keep n constant the autocorrelation procedure is refined by dividing each clone into three segments: a front section ~from fingertip to a point proximal to the distal joint); a middle section (spanning the middle phalanx); and a back section (from midfinger to the web). Each clone section is shifted separately along the parent profile, generating three separate autocorrelation curves called the front, middle, and back slither curves (Fig. 5). The undulations (nadirs and zeniths) in each slither curve for each finger locate the joints and the mid-shaft depressions relative to register, WO90/123~1 ~CT/US90/0t982 i3~ - 20 -and serve as markers for subsequent comparison of finger prominences.
The pattern of a slither curve, with its characteristic peaks and valleys, mimics the features of a house key. The shape of such a curve generated at enrollmQnt can be compared with its candidate counterpar~
point-for-point by uæing the coefficient of determination (r2), where a perfect match would yield a score o~ 1.000. A
composite of r2 values for all four f ingers is called Slither r2 and ranges in value from 0.000 to 1.000. A
composite record of three candidate slither curves for all fingers is called a REY. The corresponding record $rom previously stored slither curves (the reference record) is called a LOCR.
As with a conventional lock and key, the goal is to permit entry access only to an individual whose REY fits the pre-recorded reference LOCK. Both LOCK and KEY record the locations o the various shape prominences of all fingers in the form of a numerical matrix, wherein each cell ~0 records the number of millimeters from register to each nadir and zen~th.
Because a KEY is generated on the basis of a single hand insertion, it may be influenced by hand placement or pressure. Therefore, the LOCK is constructed from data based on multiple hand insertions at the time of - 21 - ~ ~
enrollment. In so doing, reliance is placed not on a single LOCK, but rather the KEY is allowed to try to fit the series of pre-recorded LOCKS.
~hen a REY is tried, the computer matches it sequentlally with each LQCX. Any differences between corresponding cell values are recorded in a series of least dif~erence matrices which are updated with each successive ~OC~. As the KEY is matched in turn against each LOCK, each cell value in the latest difference matrix is replaced by ~he smallest difference. Tbus, as the entire series of LOCKS ~the GRANDLOCK) is invoked, each cell value tends toward zero. The LDM score is the sum of all the cell values in the final Least Difference Matrix. A perfect match between REY and GRANDLOCR would result in an LDM score of 0.
A Greatest Difference Matrix is generated by the identical procedure ~i.e., matching the KEY to the GRANDLOCK), except that each cell value is successively replaced by the largest difference between KEY and LOCR .
The GDM score is the sum of all the cell values in the Greatest Difference Matrix. The sequential updates thereby cause the GDM score to increase.
It is expected that the cell differences in a Greatest Difference Matrix between the REY and the LOCR
series for a true match will not be very great even when the WO90/1~371 PCT/US90/01982 ~ 3 - 22 -last LOCR has been tried, and thus the GDM score for a true match will be substantially lower than for a false match.
The GDN score and LDM score are combined to yleld a measure of sameness between two profiles. This measure is the Congruence Index ~CI), based on the premise that for a true match the LDM score is much smaller than or a false match. CI is computed according to the following formula:
CI ~ (GDM Score - LDM Score)~GDM Score Thus~ as ~DM approaches zero, CI approaches lØ
An alternative formula for the quality of the match is: CI* = (GDM Score - LDM Score)/(GDM Score ~ LDM
Score) Discriminant III - Merit Index A discriminant termed the Merit Index counts the number of cells in the Least Difference Matrix whose values do not exceed some predetermined value (such as l mm).
These are called ok-cells, and are expressed as a fraction of the total number of cells in the matrix. Alternatively, count the number of ok-cells in the LDM and the GDM and apply a formula such as:
Merit Index = (GDM ok cells)/(LDM ok cells) Using this procedure for true match, the value of the Merit Index will approach l.0, since for a true match the values of GDM and LDM will be nearly the same; but for a 2 ~
false match the number of GDM ok cells will be fewer than the number of LDM ok cells and their ratio will be less than 1.0 .
Discriminant IV - Capture Area S While the CongruQnce Index (CI), Discriminant II, reveals the salient geometric features o the fingers by emphasizin~ and locating the distances between the phalangeal joints and the midshaft depressions, it leaves the actual shape comparisons of the slither curves to another discriminant called the Càpture Area (CA). The CA
parameter compares the shapes of the candidate's front.
middle, and back slither curves (the REY), with those of the stored reference curves (the LOCR~. Recall that the maximum y-axis value that any of these slither curves can achieve is 0. By delibe~ately displacing the three XEY slither curves in the y-direction with respect to their corresponding LOCR curves. and then gradually bringing them into best vertical register, the normalized area (A) per pair of data points will diminish until A between the curves reaches a minimum. The capture area (CA) for one finger is the average score obtained from the front, middle, and back slither curve comparisons.
The value of CA for the ith finger can be computed from the formula: CAi = ll-Ai]
WO90/12371 PCT/US90/~1982 For all four fingers the final value of CA is the product of the individual values: CA = CAl x CA2 x CA3 x CA4 A perfect match is achieved when the three curve shapes for each of the fingers are identical, so that when S the LOCK and ~EY curves are moved together, the a~eas (Ai~
captured between the mated curves are reduced to zero, and the value of CA ~ lØ
If there are several LOCKS in the GRAND LO~K, CA
is computed for each, and the average value (CA) (or the product) is considered as the final score.
Discriminant V - Non-Pair Score Another discriminant available from the LOCR and XEY matrices is called the Non-Pair Score (NPS). Matrices created from registered and trimmed profiles are compared, cell by cell, as in Discriminant II. ~owever, in the NPS
discriminant the value of each cell is not used, but only whether or not the cell is filled. There will be l0 KEY
matrices for l0 corresponding LOCR matrices. The l0 REY
matrices yield an average REY matrix derived from the average value of each of the corresponding cells. For a true match, nearly every filled cell in the average REY
matrix will have a corresponding filled cell in the each of the l0 LOCR matrices. A false match yielding an average KEY
2S made from the l0 constituent KEYS, when matched with the l0 constituent LOCRS in the GRANDLOCK will render a greater number of unpaired cells.
NPS is computed by counting the filled cells in the l0 LOCKS that do not have a mate in the average ~EY
matrix, and vice versa. The formula is: NPS ~ t l - np/p ]
where, np is the total number of unpaired cells and p is the total number of paired cells.
For a perfect match there will be no unpaired cells ~np ~ 0), whereupon NPS 3 1.0 Discriminant VI - Finger Size ~Area Ratio or Average Width Ratio) The following is an example of a discriminant which is not based on finger shape but rather on finqer size. Consider the finger profiles of a candidate to be matched to reference profiles. By summing the width values of the candidate fingers, compute the candidate area (Ac), and similarly obtain (Ar) for the reference area. By forming the ratio AC/Ar, or Ar/AC, whichever is smaller, obtain a new discriminant, the Area Ratio, which has the value of l.0 for a perfect size match, and less than l.0 where the finger sizes are poorly matched. The areas Ar can be expressed as the product of an average width Wr and a length L, and similarly Ac - WcL. The lengths (L) are the same because the corresponding finger lengths were equalized ~ a~ J'~ 26 -by the process of regiætering and trimming the profiles.
~ence the Area Ratio Ac/Ar ~ WC/wr (or Ar/AC ' Wr~Wc) becomes the ratio of Average Widths.
The list of potential discr~minants is very large.
Additional discriminants may be derived to discriminate between true and false matches. From among these, select the most promising combination which will lead to an appropriate accept/reject decision. Since each discriminant has a maximum value (for a perfect match) of l.0, multiply (or average) them to obtain a GRAND NATC~ NUMBE~ which itself has a maximum value of lØ The quest for the optimal combination of discriminants is an evolving process aimed at yielding the lowest possible Type I and Type II
error rates.
This invention combines a low error rate with a fast decision time, at a lower cost than other verification devices. ~and geometry is a socially acceptable and clean biometric measure which is virtually impossible to f orge.
Mechanically, the system is reliable and easy to operate.
Personnel data is encoded using available random access memory or magnetic stripe technology, while storage of a remotely located data base can be in various forms. The decision algorithms used in the system can be encrypted and are difficult to replicate or forge.
W O 90/12371 PC~r/US90/01982 2 ~ 9 3 The typical embodiment described herein shows a linear CCD (charge couple device~ sensor array in the scanning camera's image plane across which the image of the fingers is swept by means of a moving mirror. One alternative would substitute a two-dimensional CCD high resolution sensor array which, although more costly, would eliminate the need for a scanning mirror and would simplify the mechanical design. Also, the embodiment described above uses a panel backlighted by incandescent or fluorescent light sources. A high-powered flat electroluminescent panel could be used, in which case the CCD sensors must be spectrally matched to the quality of the light. Further, instead of backlighting to create the silhouette image, the panel on which the hand rests could be of a dark color while illumination of the fingers could be accomplished from above via a half-silvered mirror which would also direct the hand image to the scanning camera.
Further, width profiles hava been described as forming the basis from which the various discriminants have been derived. The finger width profiles have been described as being generated from left and right edge silhouettes of the finger images. Thus, there are eight edge silhouettes for the four fingers of each hand which may also be treated independently in the same manner as width profiles to W O 90/12371 P~r/US90/01982 provide means for testing the validity of a candidate's request ~or access.
While the method herein described, and the form of apparatus for carrying this method into effect, constitute preferred embodiments of this invention, it is to be understood that the invention is not limited to this precise method and form of apparatus, and that changes may be made in either without departing from the scope of the invention, which is de~ined in the appended claim.
What is claimed is:
` ` "`2 ~
These seven steps can be comblned into onQ simple operation. First, let ~ be the linear transformat~on on ~2 that rotates the flnger around the origln (0,0), 80 that the Qenter line of the finger is superimposQd on the vertlcal axis. In addition to the rotation, the flnger mu~t al80 be trsnslatQd to a standardizQd position rel~tive to the orlgln. ~RQmembQ~ that it was rotated around the origin, l~aving it in tbe proper orientat~on, but at some indeterminate posit~on on the plane~) To ~olve thls problem, T ~8 used to build another type of function. called an affine transformation. An affine transformation performs both rotation~ and tran~lations on its domain. The function S can be represented by a 2x2 matrix and the affine transformation u~ed can be represented by a 3x3 matrix a~-rll rl2 o t~r~ ~ r21 r22 o (t~r] could be thou~ht of as a 0 0 1 rotation matrix) w~erQ, ~T~ ~ ¦ rll rl2 ¦
I r21 r22 1 .
The above affine transformation performs only the transformAtion T, and not a translation. The mAtrix of an affine transformatlon to perform a s~mple translatlon would look l~ke:
1 o tx o o 2 ~
where, tx - the trans~ation performed along the x axis.
ty = the translation along the y axis.
The matrix of an affine transformation that will change coordinate systems would look like:
dx 0 0 ~c~ ' 0 dy 0 O O 1 ~
It is a nice property of matrices land of linear ` lO and affine transformations) that instead of performing steps l through 7 one at a time, one can combine many of these steps into one function (matrix). For this problem it is effectivè to find the center line first (before changing the coordinates of all the data points to hundredths of mm) and then build the rotation transformation A whose matrix is described above. Using this and other matrices, perform the following operations ~order is significant; matrices do not in general commute):
l) Change coordinate systems to hundredths of mm.
2) Translate the center of mass of the finger to the origin.
3) Rotate finger around origin (now center of mass).
4) Translate finger along the vertical axis so 2S that the lowest left-right pair lies on the horizontal axis.
2 ~
The matrix representation of this process would look like:
tF~] ~ 2][~,][At~A~ F~
Where the matrices of the affine transformatlons look like:
¦ dxl o 0 1 ¦ l 0 txl I
O O 1 I [A~O 1 tyl I
rll rl2 0 1 1 l 0 0 ~,] ' r21 r22 0 ~t2] - O 1 tY2 where, dxl = horizontal dimension of a pixel in hundredths of mm.
dyl = vertical dimension of a pixel in hundredths of mm.
txl 5 -(center of the fingers mass along horizontal axis).
tyl - -(center of the fingers mass along vertical axis).
rll = r22 = m/mag_m.
rl2~ = -r21 = -l/mag m.
m = slope of finqer center line.
mag m - +/- sqrt(m ~ m ~ l) Note: If m < 0 then make mag m < 0 tY2 ~ distance lin hundredths of mm) to shift the finger along the vertical axis tas in 4 above).
The data points of the finger F are represented in the matrix:
1 fxl fx2 fx3 ... fxn ¦
[F~ = fy1 fY2 fY3 fYn 1 1 1 ... 1 2 ~ 3 and each column vector (fxl,~y~ is a point on ~he S finger P.
The eguation:
lFro~m] ~ tAt2~ tA~] [Atl] tAcl] tF] -lO can be reduced to a single matrix multiplication, lF~ ][F]-where [A ] - [At2][A~][Atl][A&l].
It is desirable to calculate lA*] before working on F, for doing so greatly reduces the number of calculat~ons necessary for the computation.
After applyi*g A* to F, change the vertical coordinates from hundredths of mm to integer format mm.
Tbis obviously involves a loss of precision. To diminish the impact of this look at each mm section and average the horizontal coordinates of the points (left and right side separately, of course) that lie within the section to get the horizontal coordinate of the point at that section.
WO90/t237t PCT/US90/01982 ~ - 16 -2`~
After this reduction in resolution, data is checked for discontinuities.
The following items constitute a description of various discriminants and their derivations.
Discriminant I - Width The simplest discriminant is the coefficient of determination (r2) wherein corresponding slice widths between the reference and the candidate profiles are compared. Because the candidate's fingers may not be placed in the same positions on the panel as the reference fingers l~ a means must be followed to bring the two profiles into register so that they can be compared. This is done by shifting the profiles with respect to each other one slice at a time while testing the value of r2 at each shift until it reaches a maximum. At perfect match between identical profiles, the value of r2 ~ l.000. The point of best alignment between any two profiles is referred to as register. The numerical value of r2 at register is a meaæure of sameness of the profiles and is called Profile r2 W~90/t2371 PCT/US90/01982 - 17 _ 2~ 3 Discriminant II - Shape Prominences The first step is to locate and record tha shape prominences (joints and mid-shaft depressions) of the fingers as revealed by the reference profiles. The technique to do this is called autocorrelation. The second step is to do the same for the candidate profiles. ~he third step is to compare the locations of the correspond~ng prominences between candidate and reference autocorrelations. To extract these salient shape features from a profile, the autocorrelation procedure compares an original parent profile with itself in the form of a clone.
This has the effect of exaggerating the subtle prominences at the joints and depressions at the midshafts of the fingers.
An advantage of the autocorrelation procedure is that it forgives or accommodates any changes in the fingernail length, gross finger shape due to edema, or hand pressure, as long as such changes do not obliterate joint features altogether. Prominences located by autocorrelation àlso permit long-term identification despite normal variation in finger shapes due to aging. ~owever, it is recognized that in some persons with severe. progressive osteodegenerative disease such as arthritis, the reference profiles may have to be updated periodically.
WOgO/123~1 PCT/US90/01982 2 a~ aS~3 - 18 -The following is a description of the autocorrelation procedure which is carried out with the reference profiles and with the candidate profiles to be compared. The basic idea is described irst and a modified form is described thereafter.
A clone is made of an original (parent) profile, and an autocorrelation match curve is generated by measuring successive values of r2 as the clone profile is shifted stepwise relative to the parent profile by a distance at least as great as the separation between distal ànd proximal joints.
When the clone is in exact register with the parent profile, a peak is noted on the autocorrelation graph where the value of r2 ~ 1~000. Register, then, becomes a reference point to which prominent features of the autocorrelation curve can be referred.
Shifting the clone in l-mm intervals along the length of the parent profile, causes the value of r2 to decline as the two profiles become progressively further out of register. Minimum r2 (nadir) occurs when the distal prominence of the clone aligns with the midshaft depression of the parent. The number of l-mm intervals from register re~uired to reach thiQ nadir, determines the distance between the distal interphalangeal joint and the midshaft of 2~a~
the middle phalanx, and this value (in mm) is recorded for future reference.
Shifting the clone past the nadir to the point where the distal joint prominence of the clone lines up with the middle ~nterphalangeal p~ominence o~ the parent, locates a pseudo-match (zenith), a local maximum in the value of r2.
The number of l-mm steps from register to this zenith i8 the distance between the distal and middle interphalangeal joints. Further shifting ~misalignment) may locate yet another nadir relative to register. The number of matchable points (n) declines with each shift away from register, and this undesirable condition introduces some artifacts in the autocorrelation curve.
So, to keep n constant the autocorrelation procedure is refined by dividing each clone into three segments: a front section ~from fingertip to a point proximal to the distal joint); a middle section (spanning the middle phalanx); and a back section (from midfinger to the web). Each clone section is shifted separately along the parent profile, generating three separate autocorrelation curves called the front, middle, and back slither curves (Fig. 5). The undulations (nadirs and zeniths) in each slither curve for each finger locate the joints and the mid-shaft depressions relative to register, WO90/123~1 ~CT/US90/0t982 i3~ - 20 -and serve as markers for subsequent comparison of finger prominences.
The pattern of a slither curve, with its characteristic peaks and valleys, mimics the features of a house key. The shape of such a curve generated at enrollmQnt can be compared with its candidate counterpar~
point-for-point by uæing the coefficient of determination (r2), where a perfect match would yield a score o~ 1.000. A
composite of r2 values for all four f ingers is called Slither r2 and ranges in value from 0.000 to 1.000. A
composite record of three candidate slither curves for all fingers is called a REY. The corresponding record $rom previously stored slither curves (the reference record) is called a LOCR.
As with a conventional lock and key, the goal is to permit entry access only to an individual whose REY fits the pre-recorded reference LOCK. Both LOCK and KEY record the locations o the various shape prominences of all fingers in the form of a numerical matrix, wherein each cell ~0 records the number of millimeters from register to each nadir and zen~th.
Because a KEY is generated on the basis of a single hand insertion, it may be influenced by hand placement or pressure. Therefore, the LOCK is constructed from data based on multiple hand insertions at the time of - 21 - ~ ~
enrollment. In so doing, reliance is placed not on a single LOCK, but rather the KEY is allowed to try to fit the series of pre-recorded LOCKS.
~hen a REY is tried, the computer matches it sequentlally with each LQCX. Any differences between corresponding cell values are recorded in a series of least dif~erence matrices which are updated with each successive ~OC~. As the KEY is matched in turn against each LOCK, each cell value in the latest difference matrix is replaced by ~he smallest difference. Tbus, as the entire series of LOCKS ~the GRANDLOCK) is invoked, each cell value tends toward zero. The LDM score is the sum of all the cell values in the final Least Difference Matrix. A perfect match between REY and GRANDLOCR would result in an LDM score of 0.
A Greatest Difference Matrix is generated by the identical procedure ~i.e., matching the KEY to the GRANDLOCK), except that each cell value is successively replaced by the largest difference between KEY and LOCR .
The GDM score is the sum of all the cell values in the Greatest Difference Matrix. The sequential updates thereby cause the GDM score to increase.
It is expected that the cell differences in a Greatest Difference Matrix between the REY and the LOCR
series for a true match will not be very great even when the WO90/1~371 PCT/US90/01982 ~ 3 - 22 -last LOCR has been tried, and thus the GDM score for a true match will be substantially lower than for a false match.
The GDN score and LDM score are combined to yleld a measure of sameness between two profiles. This measure is the Congruence Index ~CI), based on the premise that for a true match the LDM score is much smaller than or a false match. CI is computed according to the following formula:
CI ~ (GDM Score - LDM Score)~GDM Score Thus~ as ~DM approaches zero, CI approaches lØ
An alternative formula for the quality of the match is: CI* = (GDM Score - LDM Score)/(GDM Score ~ LDM
Score) Discriminant III - Merit Index A discriminant termed the Merit Index counts the number of cells in the Least Difference Matrix whose values do not exceed some predetermined value (such as l mm).
These are called ok-cells, and are expressed as a fraction of the total number of cells in the matrix. Alternatively, count the number of ok-cells in the LDM and the GDM and apply a formula such as:
Merit Index = (GDM ok cells)/(LDM ok cells) Using this procedure for true match, the value of the Merit Index will approach l.0, since for a true match the values of GDM and LDM will be nearly the same; but for a 2 ~
false match the number of GDM ok cells will be fewer than the number of LDM ok cells and their ratio will be less than 1.0 .
Discriminant IV - Capture Area S While the CongruQnce Index (CI), Discriminant II, reveals the salient geometric features o the fingers by emphasizin~ and locating the distances between the phalangeal joints and the midshaft depressions, it leaves the actual shape comparisons of the slither curves to another discriminant called the Càpture Area (CA). The CA
parameter compares the shapes of the candidate's front.
middle, and back slither curves (the REY), with those of the stored reference curves (the LOCR~. Recall that the maximum y-axis value that any of these slither curves can achieve is 0. By delibe~ately displacing the three XEY slither curves in the y-direction with respect to their corresponding LOCR curves. and then gradually bringing them into best vertical register, the normalized area (A) per pair of data points will diminish until A between the curves reaches a minimum. The capture area (CA) for one finger is the average score obtained from the front, middle, and back slither curve comparisons.
The value of CA for the ith finger can be computed from the formula: CAi = ll-Ai]
WO90/12371 PCT/US90/~1982 For all four fingers the final value of CA is the product of the individual values: CA = CAl x CA2 x CA3 x CA4 A perfect match is achieved when the three curve shapes for each of the fingers are identical, so that when S the LOCK and ~EY curves are moved together, the a~eas (Ai~
captured between the mated curves are reduced to zero, and the value of CA ~ lØ
If there are several LOCKS in the GRAND LO~K, CA
is computed for each, and the average value (CA) (or the product) is considered as the final score.
Discriminant V - Non-Pair Score Another discriminant available from the LOCR and XEY matrices is called the Non-Pair Score (NPS). Matrices created from registered and trimmed profiles are compared, cell by cell, as in Discriminant II. ~owever, in the NPS
discriminant the value of each cell is not used, but only whether or not the cell is filled. There will be l0 KEY
matrices for l0 corresponding LOCR matrices. The l0 REY
matrices yield an average REY matrix derived from the average value of each of the corresponding cells. For a true match, nearly every filled cell in the average REY
matrix will have a corresponding filled cell in the each of the l0 LOCR matrices. A false match yielding an average KEY
2S made from the l0 constituent KEYS, when matched with the l0 constituent LOCRS in the GRANDLOCK will render a greater number of unpaired cells.
NPS is computed by counting the filled cells in the l0 LOCKS that do not have a mate in the average ~EY
matrix, and vice versa. The formula is: NPS ~ t l - np/p ]
where, np is the total number of unpaired cells and p is the total number of paired cells.
For a perfect match there will be no unpaired cells ~np ~ 0), whereupon NPS 3 1.0 Discriminant VI - Finger Size ~Area Ratio or Average Width Ratio) The following is an example of a discriminant which is not based on finger shape but rather on finqer size. Consider the finger profiles of a candidate to be matched to reference profiles. By summing the width values of the candidate fingers, compute the candidate area (Ac), and similarly obtain (Ar) for the reference area. By forming the ratio AC/Ar, or Ar/AC, whichever is smaller, obtain a new discriminant, the Area Ratio, which has the value of l.0 for a perfect size match, and less than l.0 where the finger sizes are poorly matched. The areas Ar can be expressed as the product of an average width Wr and a length L, and similarly Ac - WcL. The lengths (L) are the same because the corresponding finger lengths were equalized ~ a~ J'~ 26 -by the process of regiætering and trimming the profiles.
~ence the Area Ratio Ac/Ar ~ WC/wr (or Ar/AC ' Wr~Wc) becomes the ratio of Average Widths.
The list of potential discr~minants is very large.
Additional discriminants may be derived to discriminate between true and false matches. From among these, select the most promising combination which will lead to an appropriate accept/reject decision. Since each discriminant has a maximum value (for a perfect match) of l.0, multiply (or average) them to obtain a GRAND NATC~ NUMBE~ which itself has a maximum value of lØ The quest for the optimal combination of discriminants is an evolving process aimed at yielding the lowest possible Type I and Type II
error rates.
This invention combines a low error rate with a fast decision time, at a lower cost than other verification devices. ~and geometry is a socially acceptable and clean biometric measure which is virtually impossible to f orge.
Mechanically, the system is reliable and easy to operate.
Personnel data is encoded using available random access memory or magnetic stripe technology, while storage of a remotely located data base can be in various forms. The decision algorithms used in the system can be encrypted and are difficult to replicate or forge.
W O 90/12371 PC~r/US90/01982 2 ~ 9 3 The typical embodiment described herein shows a linear CCD (charge couple device~ sensor array in the scanning camera's image plane across which the image of the fingers is swept by means of a moving mirror. One alternative would substitute a two-dimensional CCD high resolution sensor array which, although more costly, would eliminate the need for a scanning mirror and would simplify the mechanical design. Also, the embodiment described above uses a panel backlighted by incandescent or fluorescent light sources. A high-powered flat electroluminescent panel could be used, in which case the CCD sensors must be spectrally matched to the quality of the light. Further, instead of backlighting to create the silhouette image, the panel on which the hand rests could be of a dark color while illumination of the fingers could be accomplished from above via a half-silvered mirror which would also direct the hand image to the scanning camera.
Further, width profiles hava been described as forming the basis from which the various discriminants have been derived. The finger width profiles have been described as being generated from left and right edge silhouettes of the finger images. Thus, there are eight edge silhouettes for the four fingers of each hand which may also be treated independently in the same manner as width profiles to W O 90/12371 P~r/US90/01982 provide means for testing the validity of a candidate's request ~or access.
While the method herein described, and the form of apparatus for carrying this method into effect, constitute preferred embodiments of this invention, it is to be understood that the invention is not limited to this precise method and form of apparatus, and that changes may be made in either without departing from the scope of the invention, which is de~ined in the appended claim.
What is claimed is:
Claims (10)
1. A method of identifying an individual based on the profile of one or more fingers on a hand, the method comprising the steps of creating a reference template by a) determining the location of the edges of each finger separately, b) establishing a central axis for each finger by reference to the edges thereof, c) rotating the central axis of each finger until it is normal to a common reference line, d) determining the finger profile by measuring the distance of each finger edge from its central axis at predetermined locations parallel to said reference line, and e) recording said finger profile information as a template; and thereafter determining a candidate's finger profile by following steps a) to d) above, comparing the candidate's finger profile with the reference template, determining whether the correspondence between the candidate's finger profile and the reference template meets certain predetermined criteria, and indicating acceptance or rejection of the candidate.
2. The method of claim l further comprising the method steps of cloning a section of each finger profile, progressively displacing the clone of each finger profile relative to the original finger profile along its axis to develop a shape curve, determining the distances between joints of each finger profile by reference to the shape curve, comparing the shape information with corresponding data obtained at the time of enrollment.
3. A method of identifying an individual based on the profile of one or more fingers, the method comprising the steps of determining the location of the edges of each finger separately thereby to generate a set of finger profiles, establishing a central axis for each finger profile by comparing the edges thereof, rotating the central axis of each finger profile until it is normal to a reference line.
cloning each finger profile, displacing the clone of each finger profile to original finger profile to determine location of joints, determining the distances between joints of each finger profile, and comparing the joint distances with corresponding hand profile data obtained at the time of enrollment.
cloning each finger profile, displacing the clone of each finger profile to original finger profile to determine location of joints, determining the distances between joints of each finger profile, and comparing the joint distances with corresponding hand profile data obtained at the time of enrollment.
4. The method of claim 3 further comprising the steps of recording both right and left hand finger profile data at the time of initial enrollment, determining whether the right or left hand is presented for comparison by analyzing lengths of individual finger elements, and retrieve the appropriate template for comparison with the candidate's finger profile.
5. The method of claim 3 wherein the location of the edges of each finger edge is determined by placing the hand on a panel, illuminating the panel on the side opposite the hand, viewing a silhouette image of the fingers by means of a scanning video camera, and scanning the fingers of the hand generally transversely to the central axis of the fingers to determine the locations of the edges of each finger at each scan thereby to develop a finger profile.
6. An apparatus for verifying the identity of an candidate based on finger profiles including means for determining a profile of each finger by locating the edges thereof, said means including a back lighted panel and video means for sensing the location of said finger edges, means for determining each finger profile independently of its orientation on said panel, means for storing a reference template comprising the finger profiles of an individual, means for comparing the finger profiles of a candidate to said reference template, and means for indicating when the candidate's finger profiles are within predetermined limits of sameness with regard to said reference template.
7. The apparatus of claim 6 wherein said video means includes a scanning video camera.
8. The apparatus of claim 6 wherein said video means includes a photosensitive linear array, a mirror, and a motor for moving the mirror to scan the fingers placed on said back lighted panel.
9. The apparatus of claim 6 wherein said video means includes a two-dimensional photosensitive array.
10. A method of verifying the identity an individual based on the profile of one or more fingers, the method comprising the steps of determining the location of the edges of each finger separately thereby to generate a set of finger profiles, establishing a central axis for each finger profile by comparing the edges thereof, rotating the central axis of each finger profile until it is normal to a common reference line, comparing the area of selected finger sections of the rotated finger profiles at the time of enrollment to the area of corresponding finger sections of the candidate.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US337,793 | 1989-04-13 | ||
| US07/337,793 US5073950A (en) | 1989-04-13 | 1989-04-13 | Finger profile identification system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA2050893A1 true CA2050893A1 (en) | 1990-10-14 |
Family
ID=23322032
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA002050893A Abandoned CA2050893A1 (en) | 1989-04-13 | 1990-04-11 | Finger profile identification system |
Country Status (9)
| Country | Link |
|---|---|
| US (1) | US5073950A (en) |
| EP (1) | EP0467964A4 (en) |
| JP (1) | JPH04504628A (en) |
| KR (1) | KR920701925A (en) |
| AU (1) | AU640648B2 (en) |
| CA (1) | CA2050893A1 (en) |
| FI (1) | FI914840A0 (en) |
| IL (1) | IL94089A (en) |
| WO (1) | WO1990012371A1 (en) |
Families Citing this family (91)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH041869A (en) * | 1990-04-19 | 1992-01-07 | Nippon Sheet Glass Co Ltd | Image collating method |
| GB2256170A (en) * | 1991-05-02 | 1992-12-02 | William Robert Brandes | Integrated circuit card with fingerprint verification. |
| US10361802B1 (en) | 1999-02-01 | 2019-07-23 | Blanding Hovenweep, Llc | Adaptive pattern recognition based control system and method |
| US5335288A (en) * | 1992-02-10 | 1994-08-02 | Faulkner Keith W | Apparatus and method for biometric identification |
| JP3378032B2 (en) * | 1992-08-28 | 2003-02-17 | 浜松ホトニクス株式会社 | Person verification device |
| US5954583A (en) * | 1992-11-05 | 1999-09-21 | Com21 Limited | Secure access control system |
| US5594806A (en) * | 1994-06-20 | 1997-01-14 | Personnel Identification & Entry Access Control, Inc. | Knuckle profile indentity verification system |
| US7248719B2 (en) | 1994-11-28 | 2007-07-24 | Indivos Corporation | Tokenless electronic transaction system |
| US7882032B1 (en) | 1994-11-28 | 2011-02-01 | Open Invention Network, Llc | System and method for tokenless biometric authorization of electronic communications |
| US20040128249A1 (en) | 1994-11-28 | 2004-07-01 | Indivos Corporation, A Delaware Corporation | System and method for tokenless biometric electronic scrip |
| US7613659B1 (en) | 1994-11-28 | 2009-11-03 | Yt Acquisition Corporation | System and method for processing tokenless biometric electronic transmissions using an electronic rule module clearinghouse |
| US7631193B1 (en) | 1994-11-28 | 2009-12-08 | Yt Acquisition Corporation | Tokenless identification system for authorization of electronic transactions and electronic transmissions |
| US6950810B2 (en) | 1994-11-28 | 2005-09-27 | Indivos Corporation | Tokenless biometric electronic financial transactions via a third party identicator |
| US5857028A (en) * | 1996-01-11 | 1999-01-05 | Frieling; Edward | Computer access control by finger anatomy and comprehension testing |
| JP3117123B2 (en) * | 1996-05-08 | 2000-12-11 | 勉 松本 | Authentication type security system |
| US5897989A (en) * | 1996-07-23 | 1999-04-27 | Beecham; James E. | Method, apparatus and system for verification of infectious status of humans |
| AU3004297A (en) * | 1996-05-13 | 1997-12-05 | James E. Beecham | Method and apparatus for ascertaining medical conditions |
| US6198836B1 (en) * | 1996-08-27 | 2001-03-06 | A-Tronic Mgm Ag | Method and device for recognizing non-unrolled fingerprints |
| US5737439A (en) * | 1996-10-29 | 1998-04-07 | Smarttouch, Llc. | Anti-fraud biometric scanner that accurately detects blood flow |
| AU5180598A (en) * | 1996-11-12 | 1998-06-22 | Express Key, Inc. | Automated check-in/check-out system |
| WO1998044447A1 (en) * | 1997-03-31 | 1998-10-08 | Bianco James S | Low-cost biometric identification system and method |
| GB9722766D0 (en) | 1997-10-28 | 1997-12-24 | British Telecomm | Portable computers |
| US7663607B2 (en) | 2004-05-06 | 2010-02-16 | Apple Inc. | Multipoint touchscreen |
| EP1717682B1 (en) | 1998-01-26 | 2017-08-16 | Apple Inc. | Method and apparatus for integrating manual input |
| DE19806024A1 (en) | 1998-02-13 | 1999-08-19 | Siemens Nixdorf Inf Syst | Method for monitoring a device operation process and self-service device monitored with it |
| DE19850577A1 (en) * | 1998-11-03 | 2000-05-04 | Fingerpin Ag Zuerich | Finger recognition device |
| GB2351585B (en) * | 1999-06-29 | 2003-09-03 | Ncr Int Inc | Self service terminal |
| AU2781401A (en) * | 2000-01-10 | 2001-07-24 | Tarian, Llc | Device using histological and physiological biometric marker for authentication and activation |
| US8049597B1 (en) | 2000-01-10 | 2011-11-01 | Ensign Holdings, Llc | Systems and methods for securely monitoring an individual |
| US9165323B1 (en) | 2000-05-31 | 2015-10-20 | Open Innovation Network, LLC | Biometric transaction system and method |
| US7565329B2 (en) | 2000-05-31 | 2009-07-21 | Yt Acquisition Corporation | Biometric financial transaction system and method |
| US6483929B1 (en) | 2000-06-08 | 2002-11-19 | Tarian Llc | Method and apparatus for histological and physiological biometric operation and authentication |
| US7441123B2 (en) * | 2001-03-23 | 2008-10-21 | Ensign Holdings | Method and apparatus for characterizing and estimating the parameters of histological and physiological biometric markers for authentication |
| US7133792B2 (en) * | 2000-06-08 | 2006-11-07 | Ensign Holdings, Llc | Method and apparatus for calibration over time of histological and physiological biometric markers for authentication |
| US7536557B2 (en) * | 2001-03-22 | 2009-05-19 | Ensign Holdings | Method for biometric authentication through layering biometric traits |
| US7148879B2 (en) | 2000-07-06 | 2006-12-12 | At&T Corp. | Bioacoustic control system, method and apparatus |
| JP3558025B2 (en) * | 2000-09-06 | 2004-08-25 | 株式会社日立製作所 | Personal authentication device and method |
| US7627145B2 (en) | 2000-09-06 | 2009-12-01 | Hitachi, Ltd. | Personal identification device and method |
| IL141389A0 (en) * | 2001-02-12 | 2002-03-10 | Weiss Golan | A system and a method for person's identity authentication |
| US7181017B1 (en) | 2001-03-23 | 2007-02-20 | David Felsher | System and method for secure three-party communications |
| US9818136B1 (en) | 2003-02-05 | 2017-11-14 | Steven M. Hoffberg | System and method for determining contingent relevance |
| US20040172562A1 (en) * | 2003-03-01 | 2004-09-02 | Vladimir Berger | System and method for identity recognition of an individual for enabling an access to a secured system |
| US20050063569A1 (en) * | 2003-06-13 | 2005-03-24 | Charles Colbert | Method and apparatus for face recognition |
| US7760918B2 (en) * | 2003-08-06 | 2010-07-20 | Zinayida Bezvershenko | Identification of a person based on ultra-sound scan analyses of hand bone geometry |
| CA2438220C (en) * | 2003-08-06 | 2011-11-08 | Click-Into Inc. | Identification of a person based on ultra-sound scan analyses of hand bone geometry |
| RU2279129C2 (en) * | 2004-02-13 | 2006-06-27 | Футроник Технолоджис Компани Лтд | Method for recognition of papillary patterns |
| US7689005B2 (en) * | 2004-03-10 | 2010-03-30 | Yingbin Wang | Apparatus and method for biometric database management system |
| US7697026B2 (en) * | 2004-03-16 | 2010-04-13 | 3Vr Security, Inc. | Pipeline architecture for analyzing multiple video streams |
| JP4348702B2 (en) * | 2004-09-29 | 2009-10-21 | 日本電気株式会社 | Plane four-finger image determination apparatus, plane four-finger image determination method, and plane four-finger image determination program |
| US8130285B2 (en) * | 2005-04-05 | 2012-03-06 | 3Vr Security, Inc. | Automated searching for probable matches in a video surveillance system |
| US8874477B2 (en) | 2005-10-04 | 2014-10-28 | Steven Mark Hoffberg | Multifactorial optimization system and method |
| US8812239B2 (en) * | 2005-12-14 | 2014-08-19 | Ut-Battelle, Llc | Method and system for real-time analysis of biosensor data |
| RU2304806C1 (en) * | 2006-01-31 | 2007-08-20 | Общество с ограниченной ответственностью "Биометрические системы и технологии" | Method for biometric identification of a person on basis of parameters of hand |
| US8552989B2 (en) | 2006-06-09 | 2013-10-08 | Apple Inc. | Integrated display and touch screen |
| CN104965621B (en) | 2006-06-09 | 2018-06-12 | 苹果公司 | Touch screen LCD and its operating method |
| KR102125605B1 (en) | 2006-06-09 | 2020-06-22 | 애플 인크. | Touch screen liquid crystal display |
| US7660442B2 (en) * | 2006-09-01 | 2010-02-09 | Handshot, Llc | Method and system for capturing fingerprints, palm prints and hand geometry |
| US8269727B2 (en) | 2007-01-03 | 2012-09-18 | Apple Inc. | Irregular input identification |
| US8130203B2 (en) | 2007-01-03 | 2012-03-06 | Apple Inc. | Multi-touch input discrimination |
| US7855718B2 (en) | 2007-01-03 | 2010-12-21 | Apple Inc. | Multi-touch input discrimination |
| US8493330B2 (en) | 2007-01-03 | 2013-07-23 | Apple Inc. | Individual channel phase delay scheme |
| US7876310B2 (en) | 2007-01-03 | 2011-01-25 | Apple Inc. | Far-field input identification |
| US9710095B2 (en) | 2007-01-05 | 2017-07-18 | Apple Inc. | Touch screen stack-ups |
| US7844915B2 (en) | 2007-01-07 | 2010-11-30 | Apple Inc. | Application programming interfaces for scrolling operations |
| US8479254B2 (en) | 2007-03-16 | 2013-07-02 | Apple Inc. | Credential categorization |
| US8151324B2 (en) | 2007-03-16 | 2012-04-03 | Lloyd Leon Burch | Remotable information cards |
| US20090204542A1 (en) * | 2008-02-11 | 2009-08-13 | Novell, Inc. | Privately sharing relying party reputation with information card selectors |
| US8150108B2 (en) * | 2008-03-17 | 2012-04-03 | Ensign Holdings, Llc | Systems and methods of identification based on biometric parameters |
| US8079069B2 (en) * | 2008-03-24 | 2011-12-13 | Oracle International Corporation | Cardspace history validator |
| US8561172B2 (en) | 2008-08-29 | 2013-10-15 | Novell Intellectual Property Holdings, Inc. | System and method for virtual information cards |
| US20100097324A1 (en) * | 2008-10-20 | 2010-04-22 | Dell Products L.P. | Parental Controls Based on Touchscreen Input |
| US8083135B2 (en) | 2009-01-12 | 2011-12-27 | Novell, Inc. | Information card overlay |
| US8632003B2 (en) | 2009-01-27 | 2014-01-21 | Novell, Inc. | Multiple persona information cards |
| US8804056B2 (en) | 2010-12-22 | 2014-08-12 | Apple Inc. | Integrated touch screens |
| US8380995B1 (en) * | 2011-11-29 | 2013-02-19 | Google Inc. | Process for login of a computing device with a touchscreen |
| US8908894B2 (en) | 2011-12-01 | 2014-12-09 | At&T Intellectual Property I, L.P. | Devices and methods for transferring data through a human body |
| US20140078303A1 (en) * | 2012-09-17 | 2014-03-20 | Jeremy Keith MATTERN | System and Method for Implementing Pass Control using an Automated Installation Entry Device |
| US9251396B2 (en) * | 2013-01-29 | 2016-02-02 | Diamond Fortress Technologies, Inc. | Touchless fingerprinting acquisition and processing application for mobile devices |
| US10108984B2 (en) | 2013-10-29 | 2018-10-23 | At&T Intellectual Property I, L.P. | Detecting body language via bone conduction |
| US9594433B2 (en) | 2013-11-05 | 2017-03-14 | At&T Intellectual Property I, L.P. | Gesture-based controls via bone conduction |
| US10678322B2 (en) | 2013-11-18 | 2020-06-09 | At&T Intellectual Property I, L.P. | Pressure sensing via bone conduction |
| US9349280B2 (en) | 2013-11-18 | 2016-05-24 | At&T Intellectual Property I, L.P. | Disrupting bone conduction signals |
| US9715774B2 (en) | 2013-11-19 | 2017-07-25 | At&T Intellectual Property I, L.P. | Authenticating a user on behalf of another user based upon a unique body signature determined through bone conduction signals |
| US9405892B2 (en) | 2013-11-26 | 2016-08-02 | At&T Intellectual Property I, L.P. | Preventing spoofing attacks for bone conduction applications |
| US9589482B2 (en) | 2014-09-10 | 2017-03-07 | At&T Intellectual Property I, L.P. | Bone conduction tags |
| US10045732B2 (en) | 2014-09-10 | 2018-08-14 | At&T Intellectual Property I, L.P. | Measuring muscle exertion using bone conduction |
| US9582071B2 (en) | 2014-09-10 | 2017-02-28 | At&T Intellectual Property I, L.P. | Device hold determination using bone conduction |
| US9882992B2 (en) | 2014-09-10 | 2018-01-30 | At&T Intellectual Property I, L.P. | Data session handoff using bone conduction |
| US9600079B2 (en) | 2014-10-15 | 2017-03-21 | At&T Intellectual Property I, L.P. | Surface determination via bone conduction |
| JP6992367B2 (en) * | 2017-09-27 | 2022-01-13 | カシオ計算機株式会社 | Nail contour detection device, nail contour detection method and nail contour detection program |
| US10831316B2 (en) | 2018-07-26 | 2020-11-10 | At&T Intellectual Property I, L.P. | Surface interface |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3576537A (en) * | 1968-12-05 | 1971-04-27 | Polaroid Corp | Hand id system |
| US3648240A (en) * | 1970-01-15 | 1972-03-07 | Identification Corp | Personnel identification apparatus |
| US3746429A (en) * | 1971-08-05 | 1973-07-17 | Recognition Systems | Linear variable light attenuator |
| US3729634A (en) * | 1971-09-20 | 1973-04-24 | Recognition Systems | Automatic beam ratio control system for holography |
| US3937580A (en) * | 1974-07-11 | 1976-02-10 | Recognition Systems, Inc. | Electro-optical method for measuring gaps and lines |
| US3944978A (en) * | 1974-09-09 | 1976-03-16 | Recognition Systems, Inc. | Electro-optical method and apparatus for making identifications |
| US4107775A (en) * | 1976-03-16 | 1978-08-15 | Novar Electronics Corporation | Human body comparison using driving point parameters |
| JPS5487251A (en) * | 1977-12-23 | 1979-07-11 | Toshiba Corp | Personal discriminator |
| US4204772A (en) * | 1978-09-25 | 1980-05-27 | Recognition Systems, Inc. | Optical measuring system |
| AU8516382A (en) * | 1981-09-21 | 1983-03-31 | Stellar Systems, Inc. | Personnel identification by finger characteristics |
| US4586441A (en) * | 1982-06-08 | 1986-05-06 | Related Energy & Security Systems, Inc. | Security system for selectively allowing passage from a non-secure region to a secure region |
| JPS6029135A (en) * | 1983-07-25 | 1985-02-14 | 三菱電機株式会社 | Indivisual discriminator |
| US4736293A (en) * | 1984-04-11 | 1988-04-05 | American Telephone And Telegraph Company, At&T Bell Laboratories | Interleaved set-associative memory |
| US4720869A (en) * | 1986-02-18 | 1988-01-19 | International Business Machines Corporation | Hand dimension verification |
-
1989
- 1989-04-13 US US07/337,793 patent/US5073950A/en not_active Expired - Fee Related
-
1990
- 1990-04-11 AU AU55243/90A patent/AU640648B2/en not_active Ceased
- 1990-04-11 KR KR1019910701339A patent/KR920701925A/en not_active Application Discontinuation
- 1990-04-11 CA CA002050893A patent/CA2050893A1/en not_active Abandoned
- 1990-04-11 EP EP19900906625 patent/EP0467964A4/en not_active Ceased
- 1990-04-11 JP JP2506477A patent/JPH04504628A/en active Pending
- 1990-04-11 WO PCT/US1990/001982 patent/WO1990012371A1/en not_active Application Discontinuation
- 1990-04-13 IL IL9408990A patent/IL94089A/en not_active IP Right Cessation
-
1991
- 1991-10-14 FI FI914840A patent/FI914840A0/en not_active Application Discontinuation
Also Published As
| Publication number | Publication date |
|---|---|
| AU5524390A (en) | 1990-11-05 |
| IL94089A (en) | 1994-11-28 |
| WO1990012371A1 (en) | 1990-10-18 |
| EP0467964A4 (en) | 1992-09-23 |
| EP0467964A1 (en) | 1992-01-29 |
| IL94089A0 (en) | 1991-01-31 |
| AU640648B2 (en) | 1993-09-02 |
| US5073950A (en) | 1991-12-17 |
| KR920701925A (en) | 1992-08-12 |
| FI914840A0 (en) | 1991-10-14 |
| JPH04504628A (en) | 1992-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US5073950A (en) | Finger profile identification system | |
| US4805223A (en) | Skin-pattern recognition method and device | |
| US4896363A (en) | Apparatus and method for matching image characteristics such as fingerprint minutiae | |
| CA2256672C (en) | Biometric recognition using a classification neural network | |
| US4947443A (en) | Method and apparatus for verifying identity | |
| US5717777A (en) | Longest line method and apparatus for fingerprint alignment | |
| US5598474A (en) | Process for encrypting a fingerprint onto an I.D. card | |
| US20090161925A1 (en) | Method for acquiring the shape of the iris of an eye | |
| US20080002860A1 (en) | Recognition method using hand biometrics with anti-counterfeiting | |
| Joshi et al. | Computer-vision-based approach to personal identification using finger crease pattern | |
| US5261008A (en) | Fingerprint verification method | |
| EP0715275B1 (en) | Automatic fingerprint identification system including processes and apparatus for matching fingerprints | |
| Dass et al. | Fingerprint-based recognition | |
| WO2002009024A1 (en) | Identity systems | |
| EP0266388A1 (en) | Identity verification | |
| EP1317730B1 (en) | Device and method for fingerprints supervision | |
| van Renesse | 3DAS: a 3-dimensional-structure authentication system | |
| EP0300167A2 (en) | Apparatus and method for matching image characteristics such as fingerprint minutiae | |
| Yau et al. | On fingerprint template synthesis | |
| van Renesse | Optical inspection techniques for security instrumentation | |
| Kaur et al. | Facial Recognition Using Theory of Biometrics | |
| Kyatham et al. | VOTE CASTING AUTHENTICATION USING BIOMETRICS | |
| JPS61217878A (en) | Method and device for personal collation | |
| IE873077L (en) | Verifying identity | |
| JPS63261491A (en) | Method for recognizing identification medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FZDE | Discontinued |