33
!ro$uc’s) +vervie/ Currently deployed field proven Biometric Speaker Identification and Verification Technology.

7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!ro$uc's) +vervie/

Currently deployed field proven Biometric Speaker Identification and Verification Technology.

Page 2: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(

,-.&/

! 0!,12(3,!4256,17

! 3850!9

! 87,7

! 37:

! 1!;1<=7,!;7

Page 3: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

0$>"&(3>$?&#'>"@A(B&C(5&"D-$E$FC(G$'(7&"H'>#C(

! Use of voice evidence in court.

! Identification of unknown intercepted voices

! Intelligence and crime prevention using phone conversations monitoring.

! Protection against Identity fraud

0$>"&(3>$?&#'>"@

Page 4: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

IDJ#(>@(0$>"&(3>$?&#'>"@K

0$>"&(3>$?&#'>"@

Extraction of unique characteristics of vocal tract, from voice sound waves.

Page 5: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

A12ITI+ !ro$uc')s +vervie/

IDJ#(>@(>-(#D&(L$>"&K

! ,-G$'?J#>$-(EJC&'A

! What the person is saying. The content of the speech.

! Can be automatically obtained using classical speech recognition technologies.

! 3&DJL>$'JE(EJC&'A

! How do you say itA Your accent, your emotions, your culture, your country DlanguageE, !"#$%&'!#(!)

! This is the level used by impersonators, and the level the brain mainly uses to identifya person Dand obtain additional informationE

0$>"&(3>$?&#'>"@

5M2(3,!4256,1(<8N26APhysical information embedded into voice waves coming from your vocal tract. Not modified by your emotional state, the language you use, etc..

!"#$%&'()**+(!"#$%&'()**+(

Page 6: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

M$O(0$>"&(3>$?&#'>"@(O$'P@K

! 5oice mo$el 8aka as ;voice <rin'>?

! Using a minimum of 40 seconds of net speech, a voice model is created for each individual.

! Contains information related to vocal tract characteristics. Unique and independent of language and text.

! Scoring or ;voice ma'cB>

! Using a minimum of M seconds of net speech, we compare the utterance with the model *'+$&,!*-'$*$./%&0#12$3"#$"-4"#0$!"#$/%&0#$!"#$5&0#$60&,*,7#$-!$-/$!"*!$!"#$8'9'&:'$;&-%#$belongs to the same person.

! 6&@HE#@

! Depending on the application, different techniques are used to .'&05*7-<#1 the scoresDset a scale common to all individuals and channelsE and set thresholds to make adecision.

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

0$>"&(3>$?&#'>"@

Page 7: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

0$>"&(3>$?&#'>"@A(B&C(1$?Q&#>#>L&(8.LJ-#JF&@

Driver& Increasing internationa0 terroris1 and cri1e using di44erent t56es o4 6hones8 the Internet8 and in 1u0ti60e 0anguages9

Voice is the only biometric characteristic obtained remotely

It is already available in abundance. Includes valuable information besides biometric information

Easy to integrate in existing infrastructure: Telephone networks, VoIP, etc..

0$>"&(3>$?&#'>"@

Page 8: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

Agni'io)s CniDue A<<roacB 

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

• Voice match is done independently of what the person is saying

Text Independent

• Voice print can be obtained in one language, and matched with voice sample in another language 

Cross Language capabilities

• Microphone, land line telephone, cellular phones, VoIP,..

Works with any voice channel

• Forensic Labs, Law enforcement, Intelligence.

Proven in the most demanding environments. 

0$>"&(3>$?&#'>"@

Page 9: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

8F->#>$A(R>$-&&'>-F(R'$.H"#@

G$'(0$>"&(M<7(8QQE>"J#>$-@

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

Present Voice Evidence in court

Identification of unknown voice during

police investigation

Speaker spotting to filter mass phone

interceptions.

0$>"&(3>$?&#'>"@

Page 10: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

ST(U$'&-@>"(2L>.&-"&(>-(1$H'#(M&J'>-F@(

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

Forensic Labs relying only on Phonetic and Linguistic techniques have limitations:

Impossible to handle unknown languages, difficult to handle great number of calls,

!"#$

Intercepted phone conversations are used as evidence in court. Experts

are required to perform Speaker Verification

MARKET NEED

SOLUTION

Tool for biometric voice verification designed for Forensic Experts. It

provides the precision and reliability required in court hearings

PROBLEM

3850!9

Page 11: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

7Q&JP&'(,V(>-(U$'&-@>"@

!1EJ@@>"JE(8QQ'$J"DA(<>-FH>@#>"WQD$-&#>"@! An expert in languages and phonetics compares both voice samples

(Suspect and unknown)

!7&?>JH#$?J#>"! Using computer voice analysis programs, sonograms, formants and other

technical information are compared.

!0$>"&(3>$?&#'>"@! Computer generates voice print, unique to speaker´s physical vocal tract.

! Computer compares suspect and unknown voice samples and provides a matching score.

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

3850!9

Page 12: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

‘A$$e$ 5alue) of 5oice Giome'rics in Horensics

! 7Q&JP&'(,V(L&'>G>"J#>$-(H@>-F(QDC@>"JE("DJ'J"#&'>@#>"@($G(#D&(L$"JE(#'J"#X(>-@#&J.($G(#D&(OJC(#D&(Q&'@$-(@Q&JP@(! >&%*7$!0*%!$.7#*;#/1$8'-?8#$@#*!80#/$-'$;&-%#$/&8'+$:*;#/A

! 0$>"&(3>$?&#'>"@(J..@(-&O(J-.('&E&LJ-#(>-G$'?J#>$-(#$("EJ@@>"JE(#&"D->YH&@T

! Depending on the case DLanguage, unique phonetic features, ..E it is more or less relevant.

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

3850!9

Page 13: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

3850!9A(R>$-&&'>-F(R'$.H"#(G$'(U$'&-@>"(<J%$'J#$'>&@

! 2-#'C(R'$.H"#

! Widely used in Europe, Latin America and Asia

! First working system in Spanish Police DIdentivox 2000E

! Expert reports presented in many courts globally

V'T(BZ-[&E(\=->LT(4J'%H'FX(]&'?J-C^(G$'?&'EC(#D&(P&C(&/Q&'#($G(#D&(]&'?J-(U&.&'JE(R$E>"&T(!-&($G(#D&(?$@#('&@Q&"#&.(@Q&">JE>@#@(O$'E._O>.&T

!"#$%&'#('')#*+,)-#.$'#/01234#%*.56%.,7#+8+.'6#999,)#65:'#.$%)#;<<#7%+'+=#ranging 4ro1 : to ;<== individua0 voice co16arisons999 Looking @ack 4ro1 toda5 I 1ust sa5 that without the he06 o4 the s5ste1 I wou0d have had to dec0ine 1ost o4 the cases8 es6ecia005 when the5 invo0ved 0anguages unknown to 1e 0ike 0>(%)'+'#5:#1*:?,+$99@

2008

3850!9

Page 14: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

BATVOX reduces learning curve, guiding the new user with a wizard, qualifying results and helping

interpretation of data

3850!9

Page 15: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

"BATVOW provides the expert Likelihood Rate DLRE results

"BC$-/$!"#$D/!*!#­of­the­*0!E$metric to measure evidence strength in court hearings

3850!9

Page 16: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

7H""&@@(7#$'CA(1$?%J#>-F(1$''HQ#>$-

! R'$?>-&-#(EJOC&'(>-(4JEJC@>J(OJ@(#JQ&.(%'$P&'>-F(#D&(JQQ$>-#?&-#(($G(#$Q(`H.F&@

! 0>.&$(Q$@#&.($-(N$H5H%&("'&J#&.(J(#H'?$>E(>-(4JEJC@>J-(Q$E>#>"@

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

%&'&()*&+,-+"*,./0012"*/+,-3!+#(,1)!4,-567879:),20/41#"),&+4,,,,,,,,,,support to verify the voice identity, and present evidence in court

3850!9

Page 17: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

7H""&@@(7#$'CA(((((((((((((((((((((((((((((((((((((((((((1$?%J#>-F(,-#&'-J#>$-JE(!'FJ->[&.(1'>?&

! 8E%J->J-(]J-F(J"#>-F(>-(]&'?J-C("$??>##&.(>.&-#>#C(G'JH.(#D'$HFD(#D&(#&E&QD$-&T

! Affecting more than Z00 German elder citi\ens

! ]&'?J-(Q$E>"&('&"$'.&.(?$'&(#DJ-(:X***(QD$-&("$-L&'@J#>$-@(%$#D(J?$-F(#D&(FJ-F(?&?%&'@(#D&?@&EL&@X(J-.($G(FJ-F(?&?%&'@(@O>-.E>-F(#D&>'(L>"#>?@T

! Conversations were in German, English, Albanian and Sinto. Some of the gang members spoke three different languages in various taped conversations.

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

German Forensic Experts verified the identity of the gang members *+;/';!4,*+,"<!,#0*=!,1)*+3,-567879:),20/41#")>,?*"<,@,/A,"<!,,,,,,,,,,,,

defendants pleading guilty after hearing the Forensic Audio evidence.

3850!9

Page 18: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

IDJ#a@(-&/#KA(SA(;(,.&-#>G>"J#>$-(1JQJ%>E>#>&@

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

1:1 Forensic evidence  verifications

1:N for Law enforcement identifications

87,7

Page 19: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

)T(<JO(2-G$'"&?&-#A(1'>?>-JE(,.&-#>G>"J#>$-(H@>-F(0$>"&

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

Police have a large amount of voices captured but identification was not possible.

Traditional methods not suitable for large voice database searches. Sometimes voice is the only available personal characteristic

of a criminal

Police obtains voice from unknown person involved in crime. (Bomb threat, lawful phone interception, terrorist

video). Identification is needed

MARKET NEED

SOLUTION

Tool for unknown speaker identification using voice

biometrics. Automatic biometric search easy to perform with

basic training

PROBLEM

87,7

Page 20: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

87,7A(0$>"&(3>$?&#'>"(VJ#J%J@&(G$'(<JO(2-G$'"&?&-#

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

B Deployed in 3 customers, 4 more pending this year.

B In the following years 20-40 countries will deploy ASIS

B $130M/yr in US onlyB Deployed in 25 countries

Fingerprint databases DNA databases

ASIS

Voice Databases

B $500M/yr marketB 25% Annual GrowthB Almost every country

AFIS CODIS

! AGNITIO follows the success of other proven biometric technologies

87,7

Page 21: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

87,7A(4$.HEJ'(J-.(@"JEJ%E&

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

87,7

DB Manager

GUIBrowser 1

Engine 1

DB

Task Manager

Application Server

Presentation Tier

GUIBrowser 2

GUIBrowser N

Business Logic

Logic Tier

Data Tier

Engine Manager

Engine 2

Engine N

Queue

Page 22: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

87,7A(4$.HEJ'(J-.(@"JEJ%E&

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

Browser 1

Browser 2

Browser 3

<HTTP> <Ethernet>

<HTTP>

<Ethernet><Ethernet>

<HTTP>

<Ethernet>

Engine 1

Engine 2

Engine NDB Server

<HTTP>Internet

Application Server Task Manager Engine Manager

ASIS MULTICORE

Active MQ

<Ethernet>

<Ethernet>

<Ethernet>

Page 23: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

• From 1,000 up to 100,000 voice models per minute.

SPEED

• From ^0,000 up to 200,000 voices stored DBithout 6er4or1ance degradationE• From ^ up to 1,000 simultaneous connections.

Maximum capacity

• Optimi\ed for HP and  Linux Suse• Engines: Proliant DL320 with Weon Quad cores• Task Manager: Integrity 3a00 with Itanium Dual core.

Hardware/OS

87,7

Page 24: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

87,7(R&'G$'?J-"&A(141

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

AGNITIO s forensic bench test Database.

How to measure performance on 1:N systemsA

CMC: Cumulative Matching Characteristic

Probability of your target being among the first N candidates

87,7

Page 25: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

Ius'omers are S'ar'ing 'o Tes' … an$  'o Cse i'…

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

A Latin A1erican securit5 organiDation9 Ehe 0argest :Fn voice test 6er4or1ed unti0 now9 A01ost :G8=== voices and 1ore than :9HM tria0s

!"

!#

$"

$#

1""

1 11 21 '1 (1 #1 )1 *1 !1 $1

1'+("",-./011#"23/131

4-52162!174-52#62$#74-521"62$*7

4-52829:23;/20913 <=->?>9093@

4-521 *27

4-52' $'7

4-52#2 $)7

An Euro6ean 6o0ice organiDation9 Eest with rea0 4orensic audios9 S1a00er data@ase9

87,7

Page 26: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

3850!9 87,7

R'&">@>$- Maximum accuracy in a 1:1 verification

Obtain a list of candidate. The verification can then be refined in a forensic laboratory

2''$'@ Minimi\e false positives

Minimi\e false negatives

=@J%>E>#C( For experts with an important background in forensics

For operatives with no background in audio and biometrics

7Q&&.( Not critical Critical: thousands of identifications per minute

87,7

Page 27: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

87,7A(0$>"&(,.&-#>G>"J#>$-

! <JH-"D&.(G'$?(J-C(JH#D$'>[&.(#&'?>-JEX(J-.(H@&'

! Minimum M seconds of speech

! Search can be guided Dover a subsetE or general

! 6&@HE#@

! Best N candidates, with scoring and a colour code, are shown to the user

! If user is authori\ed, original voice can be heard 

! There is a link to the Criminal Information Database , connected to other biometric systems

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

87,7

Page 28: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

:T(,;52<<,]2;12A(7Q&JP&'(7Q$##>-F

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

Traditional method are not useful for this kind of real time massive task. Speed,

performance and fully automatic implementation is required

Intelligence organizations do passive interceptions that

need to be filtered and classified. A key advantage

is to filter by key target speakers

MARKET NEED

SOLUTION

Tool for biometric speaker spotting. To be integrated into passive

interception systems. Detects when specific target speakers are

involved in a conversation

PROBLEM

37:

Page 29: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

37:A(B&C(G>E#&'>-F(#$$E(G$'(,-#&EE>F&-"&(8F&-">&@

! 7R28B26(V25215,!;(,;(4877(52<2RM!;2(,;52612R5,!;7! Spot a particular speaker in a continuous flow of telephone conversations

! F&567#5#'!$DG#H:&0+E$/6&!!-'4$*'+$&!"#0$DI0!-@-%-*7$J'!#77-4#'%#E$@-7!#0-'4$/H/!#5/A

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

BS3 drastically reduces time spent scanning voice in mass voice interception systems.

37:

Page 30: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

37:

Page 31: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

BATVOW

ASIS BS3

•AFIS•CODIS•IBIS•Central Criminal Data

Other Identification Databases

Selected Audios

Data crunching

Key word spotting

BS3

37:

Page 32: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

!"#$%&'()**+(A12ITI+ !ro$uc')s +vervie/

3850!9 87,7 37:

R'&">@>$- Maximum accuracy in a 1:1 verification

Obtain a list of candidate. The verification can then be refined in a forensic laboratory

Obtain a list of candidates calls where the target is speaking.

2''$'@ Minimi\e false positives

Minimi\e false negatives

Minimi\e falsenegatives

=@J%>E>#C( For experts with an important background in forensics

For operatives with no background in audio and biometrics

For Intelligence Agencies doing criminal prevention.

7Q&&.( Not critical Critical: thousands of identifications per minute

Critical: Several calls per minute.

Page 33: 7 $# .-.24&.,%) 8$29%,',:( · PROBL EM 3850 !9. 7Q&J P&'( ,V(>-( U$' &-@>"@ !1EJ@ @>"J E(8QQ'$ J" DA (< >- FH>@#> "W QD$ -&#> "@! An ex pe rt in lang ua ge

5DJ-P@(G$'(C$H'(J##&-#>$-T

U$'(GH'#D&'(>-G$'?J#>$-X(QE&J@&("$-#J"#A

Emilio Martdne\  CEO

emartine\eagnitio.es

Emmanuel Derome VP HLS Business Development

ederomeeagnitio.es

Christian Moreno HLS Business Development

chmorenoeagnitio.es

Beatri\ Gon\fle\ Siggen\a HLS Business Development

bgon\ale\eagnitio.es

!"#$%&'()**+(