1、American National StandardDeveloped byfor Information Technology Fusion Information Formatfor Data InterchangeANSI INCITS 439-2008ANSIINCITS439-2008ANSIINCITS 439-2008American National Standardfor Information Technology Fusion Information Formatfor Data InterchangeSecretariatInformation Technology I
2、ndustry CouncilApproved March 4, 2008 American National Standards Institute, Inc.AbstractThis standard specifies Fusion Information Formats (FIFs) for statistical information in support of interop-erable, modular, score-level fusion of biometric systems.Approval of an American National Standard requ
3、ires review by ANSI that therequirements for due process, consensus, and other criteria for approval havebeen met by the standards developer.Consensus is established when, in the judgement of the ANSI Board ofStandards Review, substantial agreement has been reached by directly andmaterially affected
4、 interests. Substantial agreement means much more thana simple majority, but not necessarily unanimity. Consensus requires that allviews and objections be considered, and that a concerted effort be madetowards their resolution.The use of American National Standards is completely voluntary; theirexis
5、tence does not in any respect preclude anyone, whether he has approvedthe standards or not, from manufacturing, marketing, purchasing, or usingproducts, processes, or procedures not conforming to the standards.The American National Standards Institute does not develop standards andwill in no circums
6、tances give an interpretation of any American NationalStandard. Moreover, no person shall have the right or authority to issue aninterpretation of an American National Standard in the name of the AmericanNational Standards Institute. Requests for interpretations should beaddressed to the secretariat
7、 or sponsor whose name appears on the titlepage of this standard.CAUTION NOTICE: This American National Standard may be revised orwithdrawn at any time. The procedures of the American National StandardsInstitute require that action be taken periodically to reaffirm, revise, orwithdraw this standard.
8、 Purchasers of American National Standards mayreceive current information on all standards by calling or writing the AmericanNational Standards Institute.American National StandardPublished byAmerican National Standards Institute, Inc.25 West 43rd Street, New York, NY 10036Copyright 2008 by Informat
9、ion Technology Industry Council (ITI)All rights reserved.No part of this publication may be reproduced in anyform, in an electronic retrieval system or otherwise,without prior written permission of ITI, 1250 Eye Street NW, Washington, DC 20005. Printed in the United States of AmericaCAUTION: The dev
10、elopers of this standard have requested that holders of patents that may berequired for the implementation of the standard disclose such patents to the publisher. However,neither the developers nor the publisher have undertaken a patent search in order to identifywhich, if any, patents may apply to
11、this standard. As of the date of publication of this standardand following calls for the identification of patents that may be required for the implementation ofthe standard, no such claims have been made. No further patent search is conducted by the de-veloper or publisher in respect to any standar
12、d it processes. No representation is made or impliedthat licenses are not required to avoid infringement in the use of this standard.Contents Foreword .ivIntroductionix Document Overview (Informative). 1 1 Scope . 3 2 Conformance 3 3 Normative references. 3 4 Terms and definitions. 3 4.1 biometric s
13、ample (sample) (harmonized) 3 4.2 cumulative distribution function (CDF). 3 4.3 genuine score 3 4.4 impostor score . 3 4.5 location parameter . 3 4.6 probability density function (PDF) 3 4.7 scale parameter . 4 4.8 similarity score . 4 5 Abbreviations . 4 6 Fusion Information Format (FIF) 5 6.1 Over
14、view 5 6.2 Byte ordering 6 6.3 Numeric values 6 6.4 External CBEFF structure 6 6.5 Fusion header block. 7 6.5.1 General 7 6.5.2 Format identifier. 7 6.5.3 Version number . 7 6.5.4 Record length 7 6.5.5 Biometric type 7 6.5.6 Comparison subsystem Product ID. 8 6.5.7 Dataset ID 8 6.5.8 Score sense. 9
15、6.5.9 Number of type instances 9 7 Common elements. 10 7.1 General 10 7.2 Parameter kind. 10 7.3 Parameter origin 11 7.4 Distributions present 11 7.5 Pre-normalization flag 11 8 Type 1 record. 12 8.1 Purpose 12 8.2 Format 12 8.2.1 Supporting datatype - Subtype A. 12 8.2.2 Definition 12 8.2.3 Type. 1
16、2 8.2.4 Distributions present 12 8.2.5 Impostor distribution location. 12 8.2.6 Impostor distribution scale. 13 8.2.7 Genuine distribution location . 13 8.2.8 Genuine distribution scale . 13 8.3 Use case (Informative). 13 9 Type 2 record. 13 9.1 Purpose 13 9.2 Format 13 9.2.1 Supporting datatype - S
17、ubtype B. 13 9.2.2 Definition 14 i 9.2.3 Type. 14 9.2.4 Distributions present 14 9.2.5 Impostor distribution 14 9.2.6 Genuine distribution. 14 9.3 Use case (Informative). 15 10 Type 3 record . 15 10.1 Purpose 15 10.2 Format 16 10.2.1 Supporting datatype - Subtype C 16 10.2.2 Definition 16 10.2.3 Typ
18、e. 17 10.2.4 Distributions present 17 10.2.5 Impostor distribution 17 10.2.6 Genuine distribution. 17 11 Type 4 record . 17 11.1 Purpose 17 11.2 Format 17 11.2.1 Supporting datatype - Subtype D 17 11.2.2 Definition 18 11.2.3 Type. 18 11.2.4 Distributions present 18 11.2.5 Impostor distribution 18 11
19、.2.6 Genuine distribution. 18 12 References 19 Annex A Example Cumulative Distribution Functions (informative) 20 Annex B Use of pre-normalized data (informative) . 21 Annex C Source for evaluation of spline SPLINE (informative) 23 Annex D Source code for Hermite spline interpolant HERMITE (informat
20、ive) 24 ii Tables Table 1 Fusion Information Format type taxonomy . 1 Table 2 Fusion Information Format record structure 5 Table 3 Textual representation of numerical value 6 Table 4 The fusion header block 7 Table 5 CBEFF Product Identifiers 8 Table 6 Dataset identifiers 8 Table 7 Score sense codes
21、 9 Table 8 Identifiers for statistical quantities . 10 Table 9 Origins of statistical data . 11 Table 10 Distribution information present 11 Table 11 Pre-normalization codes . 11 Table 12 Subtype A format. 12 Table 13 Type 1 record format. 12 Table 14 Subtype B format. 14 Table 15 Type 2 record form
22、at. 14 Table 16 Subtype C Format . 16 Table 17 Type 3 record format. 16 Table 18 Subtype D Format . 17 Table 19 Type 4 record format. 18 Figures Figure 1 Schematic representation of Fusion Information Format usagevi Figure A.1 - Example CDFs and their spline representations. 20 Figure B.1 - Example
23、CDFs of internal similarity scores and pre-normalized scores. 21 iii Foreword (This foreword is not part of American National Standard ANSI INCITS 439-2008.)INCITS (The International Committee for Information Technology Standards) is theANSI-recognized Standards Development Organization for informat
24、ion technologywithin the United States of America. Members of INCITS are drawn from Govern-ment, Corporations, Academia and other organizations with a material interest in thework of INCITS and its Technical Committees. INCITS does not restrict membershipand attracts participants in its technical wo
25、rk from 13 different countries, and oper-ates under the rules of the American National Standards Institute.In the field of Biometrics, INCITS has established the Technical Committee M1. Stan-dards developed by this Technical Committee have reached consensus throughoutthe development process and have
26、 been thoroughly reviewed through several PublicReview processes. In addition, this American National Standard has been approvedby the INCITS Executive Board and ANSI Board of Standards Review for Publicationas an ANSI INCITS Standard.This document contains four informative annexes, all of which are
27、 not consideredpart of this standard.Requests for interpretation, suggestions for improvement or addenda, or defect re-ports are welcome. They should be sent to InterNational Committee for InformationTechnology Standards (INCITS), ITI, 1250 Eye Street, NW, Suite 200, Washington,DC 20005.This standar
28、d was processed and approved for submittal to ANSI by INCITS. Com-mittee approval of this standard does not necessarily imply that all committee mem-bers voted for its approval. At the time it approved this standard, INCITS had thefollowing members:Karen Higginbottom, ChairJennifer Garner, Secretary
29、Organization Represented Name of RepresentativeAdobe Systems, Inc Leslie BixelSteve Ziles (Alt.)AIM Global, Inc. Dan MullenCharles Biss (Alt.)Apple Computer, Inc. David MichaelElectronic Industries Alliance Edward Mikoski, Jr.David Thompson (Alt.)EMC Corporation Gary RobinsonFarance, Inc Frank Faran
30、ceTimothy Schoechle (Alt.)GS1 US Frank SharkeyJames Chronowski (Alt.)Mary Wilson (Alt.)Hewlett-Packard Company. Karen HigginbottomSteve Mills (Alt.)Scott Jameson (Alt.)IBM Corporation . Ronald F. SillettiRobert Weir (Alt.)Sandy Block (Alt.)Richard Schwerdtfeger (Alt.)IEEE . Judith GormanTerry DeCour
31、celle (Alt.)Bill Ash (Alt.)Jodi Haasz (Alt.)Bob Labelle (Alt.)ivvOrganization Represented Name of RepresentativeIntel .Philip WennblomDave Thewlis (Alt.)Jesse Walker (Alt.)Grace Wei (Alt.)Lexmark InternationalDon WrightDwight Lewis (Alt.)Paul Menard (Alt.)Microsoft CorporationJim HughesDon Stanwyck
32、(Alt.)Isabelle Valet-Harper (Alt.)National Institute of Standards for similarities, higher values indicate this. A 1 byte field with values allowed by Table 7 shall be used to record score sense. Table 7 Score sense codes Sense Value (1 byte) Distance 0 Similarity 1 NOTE The use of the word distance
33、 does not necessarily imply that values have the metric property. 6.5.9 Number of type instances The (1-byte) number of type instances block shall be the number of Type 1, 2, 3 or 4 records included in the FIF record. A record shall contain zero or one instance of each Type, but there shall always b
34、e one instance of one of them. Thus the number of type instances shall be on 1, 2, 3 or 4. EXAMPLE If both a Type 1 and Type 3 instance are present, this value will be 2. NOTE An application profile (or equivalent specification) might appropriately call-out one specific type. 9 ANSI INCITS 439-2008
35、7 Common elements 7.1 General This clause defines generic data structures to support the Type 1, 2, 3 and 4 records of this standard. This clause also includes tables that enumerate values and associates them with their respective meanings. Note that in some tables the first column is entitled “Fiel
36、d“ and contains a numeric entry. These are included only for ease of reference to the standard the values should not be included in instances of the standards binary records. 7.2 Parameter kind Table 8 gives integer values as identifiers for quantities needed to describe a distribution. Table 8 Iden
37、tifiers for statistical quantities Kind Description Notes 0 unspecified the producer neglects to specify (usually unacceptable) 1 unknown unknown quantity 2 mean 3 median 4 mode 5 minimum extreme values not true location parameters 6 maximum 7 (minimum + maximum) / 2 8 Tukeys trimean 9 generic locat
38、ion parameter 10 31 undefined 32 variance 33 standard deviation 34 median absolute deviation (MAD) Actually 1.4826 (med(abs(x-med) where the constant 1.4826 normalizes so that the expectation value of the mad is equal to the std. deviation of a normal distribution. 35 (maximum minimum) extreme value
39、s not true location parameters 36 generic scale parameter 37 skewness 38 kurtosis 39 65 undefined 66 generic parameter I for distributions parameterized by more than just location and scale 67 generic parameter II 68 95 undefined 96 cumulative distribution function as discrete pairs (xi, F(xi) for i
40、 = 1N as used in Type 297 cumulative distribution function as a B-spline representation used in Type 398 cumulative distribution function as a Hermite spline interpolant representation used in Type 499 probability density function as histogram 100 255 undefined NOTE 1 An application may well effect
41、the same transformation regardless of which kind of values the record holds, e.g. linear transforms such as (x-mean)/stdev and (x-median)/mad. 10 ANSI INCITS 439-2008 NOTE 2 The range of parameters is wide. For a specific application it will be necessary to narrowly specify these. This is properly d
42、one in an application profile, or other requirements document. 7.3 Parameter origin The (1 byte) value of Table 9 shall be used to indicate the origin of the parameter. Table 9 Origins of statistical data Origin Distribution of what quantity 0 Unspecified value(s) whose origin is not divulged 1 Unkn
43、own value(s) with undetermined origin 2 Empirical value(s) estimated from experiment and sample data 3 Known a priori value(s) known by design, or theoretical considerations. For a distribution, this value would indicate it is known in closed-form. 4 254 Reserved 7.4 Distributions present The (1 byt
44、e) value shall be used to indicate which of the impostor and genuine distributions are included in a record. The allowed values are given in Table 10. This field shall contain the value 0x01, 0x02 or 0x03. When both distributions are present the impostor distribution shall precede the genuine distri
45、bution. Table 10 Distribution information present Value Distributions present 0x01 Impostor 0x02 Genuine 0x03 Impostor and Genuine 0x04 - 0xFF Reserved 7.5 Pre-normalization flag The (1 byte) value of Table 11 shall be used to indicate whether scores from the comparison subsystem exist on an arbitra
46、ry range or have been pre-normalized. For the purposes of this standard this means that a fusion module may regard the scores produced by the comparison subsystem to be uniformly distributed on the range 0,1. This supports direct use of scores without need for interpolation. See Annex B for informat
47、ion on the meaning of this field. NOTE 1 This field is present in the Type 2, 3, or 4 records for the genuine and impostor distributions. If a comparison subsystem produces pre-normalized impostor scores, the genuine scores will not be uniform, and vice versa. Thus it will never make sense for the r
48、ecords of this standard to indicate that both the impostor and genuine scores are pre-normalized. NOTE 2 Implementations conformant to BioAPI return “FAR values“. This means that internally generated impostor scores are either naturally on 0,1 or have been normalized by the expected impostor CDF of
49、those scores. The result will be that the output impostor scores can be expected to be uniform on 0,1. Table 11 Pre-normalization codes Pre-normalization status Range of scores Value (1 byte) Not normalized 0 - N points. A function such as the Hermite spline interpolant HERMITE of clause 11 is inferior to B-spline interpolation when n N. NOTE 6 Annex C presents C+ code for the evaluation of the spline function. 16
