(2007) performance analysis for multi sensor fingerprint recognition system

51
Collaborative Project with Purdue University Since August 2006 Analysis on Analysis on Finger Preference of Users for Fi i tR iti S t Fingerprint Recognition Systems Jihyun Moon and Hale Kim, Inha Univ. Matt Young and Steve Elliott, Purdue Univ. February 1, 2007 2007 BERC Workshop

Upload: international-center-for-biometric-research

Post on 25-Jun-2015

288 views

Category:

Technology


3 download

DESCRIPTION

The increasing use of distributed authentication architecture has made interoperability of systems an important issue. Interoperability of systems re°ects the maturity of the technology and also improves confidence of users in the technology. Biometric systems are not immune to the concerns of interoperability. Interoperability of fingerprint sensors and its effect on the overall performance of the recognition sys- tem is an area of interest with a considerable amount of work directed towards it. This research analyzed effects of interoperability on error rates for fingerprint datasets captured from two optical sensors and a capacitive sensor when using a single commercially available fingerprint matching algorithm. The main aim of this research was to emulate a centralized storage and matching architecture with multiple acquisition stations. Fingerprints were collected from 44 individuals on all three sensors and interoperable False Reject Rates of less than .31% were achieved using two different enrolment strategies.

TRANSCRIPT

Page 1: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

Collaborative Project with Purdue UniversitySince August 2006

Analysis onAnalysis on Finger Preference of Users for

Fi i t R iti S tFingerprint Recognition Systems

Jihyun Moon and Hale Kim, Inha Univ.Matt Young and Steve Elliott, Purdue Univ.

February 1, 2007

2007 BERC Workshop

Page 2: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Contentsre

fere

nc

IntroductionTest Protocol

n a

nd

Pr

Data Collection Analysis Results

bit

uati

on y

Conclusion and Future Work

s o

n H

ab

An

aly

sis

A

2

Page 3: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Introductionre

fere

nc

BackgroundUser habituation is a concept and term that is widely

d ithi th bi t i i d t i ll th t f

n a

nd

Pr used within the biometric industry, especially that of

the biometric standards community, without a detailed definition or means necessary to measure

bit

uati

on y

and determine habituation.

s o

n H

ab Purpose

To assess the concept of “habituation” for biometric systems fingerprints in particular for this study

An

aly

sis systems, fingerprints in particular for this study.

Furthermore, to investigate the affects of handedness and finger preference in discussing “habituation”.A

3

g p g

Page 4: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Test Protocolre

fere

nc Expected Measurements

Demographic information PURDUE & INHA

n a

nd

Pr Handedness PURDUE & INHA

Finger preference PURDUE & INHAImage quality (NFIQ) INHA

bit

uati

on Image quality (NFIQ) INHA

Condition: Use-ness of Hand cream

Region of Overlap INHA

s o

n H

ab

g pMatching Performance (BOZORTH3) INHAStatistical analysis

An

aly

sis

ROC curves in time series for each groupROC curves for inter-group analysis (except for G4)Distributions in time seriesA

4

IQ, OL

Page 5: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Test Protocol(INHA)re

fere

nc

SensorU.Are.U4000 by DigitalPersona

n a

nd

Pr

SubjectAt least 15 persons per each group

bit

uati

on

Procedure (enrolled, attempt)Group# Week1 Week2 Week3 Week4 Week5 Week6

1 7 i 3 i 3 i 3 i 3 i 3 i

s o

n H

ab 1 7 images

3 images3 images 3 images 3 images 3 images 3 images

2 7 images3 images

3 images

An

aly

sis 3 images

3 7 images3 images

3 images

A

5

4 7 images3 images/3 images

Page 6: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Test Protocolre

fere

nc

Fingerprint Image Filename conventionSSS-F-I-W#.bmp

n a

nd

Pr Where SSS is the subject #, F is the finger # (1-3), I is the

image # (1-10 or 1-3), and W# is the week # (1-6). Purdue will use the subject #s from 1-99. Inha will use the

bit

uati

on subject #s from 100 – 199 (or greater as needed).

Restrictions

s o

n H

ab Restrictions

Role of the supervisorThe supervisor shall inform the users of the instructions con

An

aly

sis

tained in the first point in the delimitations section. However; the supervisor shall not engage in any physical assistance of any sort while the user is submitting finger images.

A

6

Page 7: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(INHA)re

fere

nc Subjects

100 120SubjectID

126 143SubjectID

Group1 Group2

n a

nd

Pr 100 120

101 121102 122105 123108 124111 125

126 143131 144132 145133 146136 147137 148

bit

uati

on 112 127

113 128114 129118

Total Subjects 19

138 149139 150140 151141 152

Total Subjects 20

s o

n H

ab

153 164 174154 165 175155 166 176

SubjectID183 193 203184 194 204185 195 205

SubjectIDGroup3 Group4

An

aly

sis 155 166 176

156 167 177157 168 178159 169 179160 170 180161 171 181

185 195 205186 196 206187 197188 198189 199190 200A

7

162 172 182163 173

Total Subjects 29

191 201192 202

Total Subjects 24

Page 8: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(BSPA)re

fere

nc Subjects

SubjectID

Group1 Group2

SubjectID

n a

nd

Pr

2 133 154 1769

SubjectID18 3019 3121 692224

SubjectID

bit

uati

on 11

12Total Subjects 10

2526

Total Subjects 10

s o

n H

ab

Group3 Group4

28 4235 44

SubjectID50 5952 60

SubjectID

An

aly

sis 35 44

36 4637 4738 4839 7140

S

52 6053 6555565758

SA

8

Total Subjects 13 Total Subjects 10

Page 9: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(INHA)re

fere

nc

ProcedureBefore adjustment

n a

nd

Pr

Group# Week1 Week2 Week3 Week4 Week5 Week6

1 10 images 3 images 3 images 3 images 3 images 3 images

2 10 images 3 images

bit

uati

on 2 10 images 3 images

3 10 images 3 images

4 10 images/3 images

s o

n H

ab

After adjustmentGroup# Week1 Week2 Week3 Week4 Week5 Week6 Week7

/3 images

An

aly

sis p

1 10 images 3 images 3 images 3 images 3 images 3 images

2 10 images 3 images

3 10 images 3 imagesA

9

3 10 images 3 images

4 10 images/3 images

Page 10: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(BSPA)re

fere

nc

ProcedureBefore adjustment

n a

nd

Pr

Group# Week1 Week2 Week3 Week4 Week5 Week6

1 10 images 3 images 3 images 3 images 3 images 3 images

2 10 images 3 images

bit

uati

on 2 10 images 3 images

3 10 images 3 images

4 10 images/3 images

s o

n H

ab

After adjustmentGroup# Week1 Week2 Week3 Week4 Week5 Week6 Week7

/3 images

An

aly

sis p

1 10 images 3 images 3 images 3 images 3 images

2 10 images 3 images

3 10 images 3 imagesA

10

3 10 images 3 images

4 10 images/3 images

Page 11: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(INHA)re

fere

nc

Fingerprint Image Filename convention

n a

nd

Pr Before adjustment: SSS-F-I-W#.bmp

After adjustment: SSS-F-II-#.bmp

File Size (Adjustment)

bit

uati

on File Size (Adjustment)

Fix size to Width 320 and Height 352

Number of Images

s o

n H

ab Group Images per subject Subjects Total

1 75 19 14252 39 20 7803 39 29 1131

An

aly

sis

4 39 24 936

A

11

Page 12: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collection(BSPA)re

fere

nc

Fingerprint Image Filename convention

n a

nd

Pr Before adjustment: SSS-F-I-W#.bmp

After adjustment: SSS-F-II-#.bmp

File Size (Adjustment)

bit

uati

on File Size (Adjustment)

Fix size to Width 320 and Height 352

Number of Images

s o

n H

ab

Group Images per subject Subjects Total1 66 10 6602 39 10 3903 39 13 507

An

aly

sis

4 39 10 390

A

12

Page 13: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collectionre

fere

nc

Environment1 – CVLab

n a

nd

Pr

bit

uati

on

s o

n H

ab

An

aly

sis

A

13

Page 14: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collectionre

fere

nc

Environment2 – Computer Room 424

n a

nd

Pr

bit

uati

on

s o

n H

ab

An

aly

sis

A

14

Page 15: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Data Collectionre

fere

nc

Forms for collecting information

n a

nd

Pr

bit

uati

on

s o

n H

ab

An

aly

sis

A

15

Page 16: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Summary on Personal Data

G 1 19Total # of subjects Gender Male Female

G 1 18 1

(persons) (persons)

n a

nd

Pr Group1 19

Group2 20Group3 29Group4 24

Group1 18 1Group2 16 4Group3 28 1Group4 24 0

(persons)93.5%

bit

uati

on Ethnicity Indian etc. Asian Black Hispanic etc. White Other

Group1 0 19 0 0 0 0Group2 0 20 0 0 0 0Group3 0 29 0 0 0 0Group4 0 24 0 0 0 0

100 0%

s o

n H

ab

Handedness Right Left AmbidextrousGroup1 16 2 1Group2 16 1 3Group3 26 2 1

(persons) <고찰>

1. 성별 구성에 있어서 남성의 비율이 월등히 큼

2. 모두 동양인임 대부분 오른손잡이임

3 남성이 대부분이기 때문에 피부 보습을 위한

100.0%

An

aly

sis

Hand cream Yes NoGroup1 6 13Group2 8 12

Used before Yes NoGroup1 3 16Group2 3 17

Group4 23 0 1

(persons) (persons)

3. 남성이 대부분이기 때문에 피부 보습을 위한조처를 하지 않는 경우가 상대적으로 많음

4. 지문 인식 시스템을 사용해본 경험이 없는사람들로 참여자를 구성하고자 하였으나, 면접후 일부에서 사용 경험이 있었던 것으로 나타남

88.0%

A

16

Group2 8 12Group3 9 20Group4 6 18

Group2 3 17Group3 4 25Group4 3 21

68.5% 85.9%

Page 17: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Group1 10Total # of subjects

(persons) (persons)

Gender Male FemaleGroup1 8 2

Summary on Personal Data

n a

nd

Pr Group1 10

Group2 10Group3 13Group4 10

(persons)86.0%

Group1 8 2Group2 8 2Group3 12 1Group4 9 1

bit

uati

on Ethnicity Indian etc. Asian Black Hispanic etc. White Other

Group1 0 1 0 0 9 0Group2 0 0 0 0 10 0Group3 0 0 0 0 13 0Group4 1 0 0 0 9 0

95 3%

s o

n H

ab

Handedness Right Left AmbidextrousGroup1 9 0 1Group2 8 1 1Group3 13 0 0

(persons) <고찰>

1. 성별 구성에 있어서 남성의 비율이 월등히 큼

2. 모두 동양인임 대부분 오른손잡이임

3 남성이 대부분이기 때문에 피부 보습을 위한

95.3%

An

aly

sis Group4 10 0 0

(persons) (persons)

3. 남성이 대부분이기 때문에 피부 보습을 위한조처를 하지 않는 경우가 상대적으로 많음

4. 지문 인식 시스템을 사용해본 경험이 없는사람들로 참여자를 구성하고자 하였으나, 면접후 일부에서 사용 경험이 있었던 것으로 나타남

93.0%

Hand cream Yes NoGroup1 3 7Group2 2 8

Used before Yes NoGroup1 6 4Group2 7 3A

17

74.4% 51.2%

Group2 2 8Group3 2 11Group4 4 6

Group2 7 3Group3 4 9Group4 5 5

Page 18: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Summary on Personal Data

Total # of subjects

n a

nd

Pr Total # of subjects

Group1 Group2 Group3 Group4

bit

uati

on Group1

21%Group426%

s o

n H

ab

Group222%

Group331%

An

aly

sis 31%

A

18

Page 19: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Summary on Personal Data

Total # of subjects

n a

nd

Pr Total # of subjects

Group1 Group2 Group3 Group4

bit

uati

on Group1

23%Group4

23%

s o

n H

ab

Group223%Group3

31%

An

aly

sis

A

19

Page 20: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Summary on Personal Data

Gender distributions of 4 groups

n a

nd

Pr Gender distributions of 4 groups

30

35

nsMale Female

bit

uati

on

15

20

25

30

r of

per

son

s o

n H

ab

0

5

10

G 1 G 2 G 3 G 4

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

20

Page 21: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Gender distributions of 4 groups

Summary on Personal Data

n a

nd

Pr Gender distributions of 4 groups

15

s

Male Female

bit

uati

on

10

er o

f pe

rson

s

s o

n H

ab

0

5

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

21

Page 22: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Summary on Personal Data

Handedness distributions of 4 groups

n a

nd

Pr Handedness distributions of 4 groups

30

nsRight Left Ambidextrous

bit

uati

on

15

20

25

r of

per

son

s o

n H

ab

0

5

10

G 1 G 2 G 3 G 4

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

22

Page 23: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Summary on Personal Data

Handedness distributions of 4 groups

n a

nd

Pr Handedness distributions of 4 groups

15

s

Right Left Ambidextrous

bit

uati

on

10

er o

f pe

rson

s

s o

n H

ab

0

5

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Gro up#

A

23

Page 24: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Summary on Personal Data

Hand cream useness distributions of 4 groups

n a

nd

Pr Hand- cream useness distributions of 4 groups

30

nsYes No

bit

uati

on

15

20

25

r of

per

son

s o

n H

ab

0

5

10

G 1 G 2 G 3 G 4

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

24

Page 25: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Summary on Personal Data

Hand cream useness distributions of 4 groups

n a

nd

Pr Hand- cream useness distributions of 4 groups

15

sYes No

bit

uati

on

10

er o

f pe

rson

s

s o

n H

ab

0

5

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

25

Page 26: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Sensor useness distributions of 4 groups

Summary on Personal Data

n a

nd

Pr Sensor useness distributions of 4 groups

30

nsYes No

bit

uati

on

15

20

25

r of

per

son

s o

n H

ab

0

5

10

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

26

Page 27: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Summary on Personal Data

Sensor useness distributions of 4 groups

n a

nd

Pr Sensor useness distributions of 4 groups

15

s

Yes No

bit

uati

on

10

r of

per

sons

s o

n H

ab

0

5

Num

be

An

aly

sis Group1 Group2 Group3 Group4

Group#

A

27

Page 28: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Resultsre

fere

nc

Summary on Environmental Data (average)Temperature Week1 Week2 Week3 Week4 Week5 Week6G 1 26 9 27 6 27 2 27 5 29 7 26 7

(°C)

n a

nd

Pr Group1 26.9 27.6 27.2 27.5 29.7 26.7

Group2 27.3 26.8Group3 28.9 27.6Group4 25.4

(%)

bit

uati

on

Moisture Week1 Week2 Week3 Week4 Week5 Week6Group1 41.5 39 35.5 40.5 38.5 49Group2 43 51.9Group3 42 29.6Group4 54.2

(%)

s o

n H

ab G oup 5

An

aly

sis

A

28

Page 29: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Finger Preference Result

Finger Preference - 1st Finger

n a

nd

Pr Finger Preference - 1st Finger

1618

Group1 Group2 Group3 Group4

bit

uati

on

8101214

eque

ncy

s o

n H

ab

0246

LL LR LM LI LT RT RI RM RR RL

Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

29

Page 30: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Finger Preference Result

Finger Preference - 1st Finger

n a

nd

Pr Finger Preference 1st Finger

12

Group1 Group2 Group3 Group4

bit

uati

on

6

8

10

eque

ncy

s o

n H

ab

0

2

4

LL LR LM LI LT RT RI RM RR RL

Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

30

Page 31: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Finger Preference Result – cont.

Finger Preference 2nd Finger

n a

nd

Pr Finger Preference - 2nd Finger

12

14

Group1 Group2 Group3 Group4

bit

uati

on

6

8

10

12

eque

ncy

s o

n H

ab

0

2

4

LL LR LM LI LT RT RI RM RR RL

Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

31

Page 32: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Finger Preperence 2nd Finger

Finger Preference Result – cont.

n a

nd

Pr Finger Preperence - 2nd Finger

6

Group1 Group2 Group3 Group4

bit

uati

on

4

eque

ncy

s o

n H

ab

0

2

LL LR LM LI LT RT RI RM RR RL

Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

32

Page 33: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Finger Preference Result – cont.

Finger Preference 3rd Finger<고찰>

n a

nd

Pr Finger Preference - 3rd Finger

12

Group1 Group2 Group3 Group4

1. 지문 영상 수집 실험 전 선택한 선호 손가락은, 사용자들의 무의식에 의한 선택에 기반한 경우가많았다고 생각됨

2. 오른손잡이가 많았던 관계로 무의식적으로오른손 세 손가락을 선택하는 경우가 많았음

bit

uati

on

6

8

10

eque

ncy

오른손 세 손가락을 선택하는 경우가 많았음

3. 특히, 첫 번째 선호 손가락은 엄지와 검지인경우가 대부분이었으며, 순번이 뒤로 갈 수록선택에 있어서 별 다른 기준이 없는 것처럼 보임

s o

n H

ab

0

2

4Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

33

Page 34: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Finger Preference 3rd Finger

Finger Preference Result – cont.

n a

nd

Pr Finger Preference - 3rd Finger

6

Group1 Group2 Group3 Group4

bit

uati

on

4

quen

cy

s o

n H

ab

0

2Fre

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

34

Page 35: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)(persons)

refe

ren

c

Finger Preference Change

After Finger Preference Change

Prefernecechange Yes No

Group1 9 10Group2 6 14Group3 17 12

(persons)

n a

nd

Pr g g

25

Finger1 Finger2 Finger3

pGroup4 8 16

bit

uati

on

10

15

20

requ

ency

s o

n H

ab

0

5

10F

An

aly

sis

LL LR LM LI LT RT RI RM RR RLFinger Index<고찰>

1. 센서 사용 후의 선호도 변화를 살펴 본 결과검지(LI, RI)로의 선호도 변화가 뚜렷해 짐.A

35

2. 양손 엄지(LT, RT) 및 검지(LI, RI)로의선호도 변화가 전체의 약 80.8%를 차지함.

Page 36: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)(persons)

refe

ren

c

Finger Preference Change(persons)

After Finger Preference Change

Preferencechange Yes No

Group1 3 7Group2 4 6Group3 5 8

n a

nd

Pr g g

78

Finger1 Finger2 Finger3

Group3 5 8Group4 8 2

bit

uati

on

3456

Freq

uenc

y

s o

n H

ab

0123

LL LR LM LI LT RT RI RM RR RL

F

An

aly

sis LL LR LM LI LT RT RI RM RR RL

Finger Index

A

36

Page 37: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Finger Preference Change – occurrence tableFinger1B f LL LR LM LI LT RT RI RM RR RL

After preference changed Finger2B f LL LR LM LI LT RT RI RM RR RL

After preference changed

n a

nd

Pr Before LL LR LM LI LT RT RI RM RR RL

LL 0 0 1 0 1 2 1 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 0 0 0 0 0 0 0 0LI 0 0 0 3 1 0 0 0 0 0LT 0 0 0 3 0 0 1 0 0 0

Before LL LR LM LI LT RT RI RM RR RLLL 0 0 0 0 0 0 0 0 0 0LR 0 0 0 1 0 0 1 0 0 0LM 0 0 0 0 0 0 0 0 0 0LI 0 0 0 1 0 1 1 0 0 0LT 1 0 0 2 1 1 2 0 0 0

bit

uati

on RT 0 0 0 1 0 0 15 1 0 0

RI 0 0 0 0 0 3 7 0 0 0RM 0 0 0 0 0 0 0 0 0 0RR 0 0 0 0 0 0 0 0 0 0RL 0 0 0 0 0 0 0 0 0 0

RT 0 0 0 2 0 1 1 0 0 0RI 0 0 1 5 1 2 2 4 0 1

RM 0 0 0 3 2 1 1 0 0 0RR 0 0 0 0 0 0 0 0 0 0RL 0 0 0 1 0 0 0 0 0 0

s o

n H

ab Finger3

Before LL LR LM LI LT RT RI RM RR RLLL 0 0 0 0 1 0 0 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 0 0 2 2 0 1 0 0LI 0 0 0 1 0 2 2 1 0 0

After preference changed AllBefore LL LR LM LI LT RT RI RM RR RL

LL 0 0 1 0 2 2 1 0 0 0LR 0 0 0 1 0 0 1 0 0 0LM 0 0 0 0 2 2 0 1 0 0LI 0 0 0 5 1 3 3 1 0 0

After preference changed

An

aly

sis LI 0 0 0 1 0 2 2 1 0 0

LT 0 0 0 1 1 0 0 1 0 0RT 0 0 1 0 1 0 1 1 0 0RI 0 0 0 1 1 2 0 1 0 0

RM 0 0 1 2 0 3 0 3 2 0RR 0 0 0 0 1 2 0 1 0 0RL 0 0 0 0 0 0 0 1 0 0

LI 0 0 0 5 1 3 3 1 0 0LT 1 0 0 6 2 1 3 1 0 0RT 0 0 1 3 1 1 17 2 0 0RI 0 0 1 6 2 7 9 5 0 1

RM 0 0 1 5 2 4 1 3 2 0RR 0 0 0 0 1 2 0 1 0 0RL 0 0 0 1 0 0 0 1 0 0A

37

RL 0 0 0 0 0 0 0 1 0 0 RL 0 0 0 1 0 0 0 1 0 0

80.8%

Page 38: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Finger1 After preference changed Finger2 After preference changed

Finger Preference Change – occurrence table

n a

nd

Pr Before LL LR LM LI LT RT RI RM RR RL

LL 0 0 0 0 0 0 0 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 0 0 0 0 0 0 0 0LI 0 0 0 0 0 0 0 0 0 0LT 0 0 0 0 1 0 0 0 0 0RT 0 0 0 0 0 2 7 0 0 0

Before LL LR LM LI LT RT RI RM RR RLLL 0 0 0 0 0 0 0 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 0 0 0 0 0 0 0 0LI 0 0 0 1 0 0 0 0 0 0LT 0 0 0 0 2 0 0 0 0 0RT 0 0 0 0 2 7 0 4 0 0

bit

uati

on RT 0 0 0 0 0 2 7 0 0 0

RI 0 0 0 0 0 0 31 0 0 0RM 0 0 0 0 0 0 0 2 0 0RR 0 0 0 0 0 0 0 0 0 0RL 0 0 0 0 0 0 0 0 0 0

RT 0 0 0 0 2 7 0 4 0 0RI 0 0 0 0 0 1 2 5 1 0

RM 0 0 0 1 0 2 0 14 0 0RR 0 0 0 0 0 0 0 0 1 0RL 0 0 0 0 0 0 0 0 0 0

Finger3 After preference changed All After preference changed

s o

n H

ab Finger3

Before LL LR LM LI LT RT RI RM RR RLLL 0 0 0 0 0 0 0 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 1 0 0 0 0 1 0 0LI 0 0 0 5 0 1 0 0 0 0LT 0 0 1 0 5 1 0 0 0 0

After preference changed AllBefore LL LR LM LI LT RT RI RM RR RL

LL 0 0 0 0 0 0 0 0 0 0LR 0 0 0 0 0 0 0 0 0 0LM 0 0 1 0 0 0 0 1 0 0LI 0 0 0 6 0 1 0 0 0 0LT 0 0 1 0 8 1 0 0 0 0

After preference changed

An

aly

sis RT 0 0 1 0 0 3 0 1 2 0

RI 0 0 0 0 0 0 0 0 0 0RM 0 0 0 3 1 1 0 7 2 0RR 0 0 0 1 0 1 0 0 5 0RL 0 0 0 0 0 0 0 0 0 0

RT 0 0 1 0 2 12 7 5 2 0RI 0 0 0 0 0 1 33 5 1 0

RM 0 0 0 4 1 3 0 23 2 0RR 0 0 0 1 0 1 0 0 6 0RL 0 0 0 0 0 0 0 0 0 0

A

38

Page 39: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(INHA)re

fere

nc

Finger Preference ChangeAll

Before LL LR LM LI LT RT RI RM RR RL TotalLL 0 0 1 0 2 2 1 0 0 0 6

After preference changed

n a

nd

Pr LL 0 0 1 0 2 2 1 0 0 0 6

LR 0 0 0 1 0 0 1 0 0 0 2LM 0 0 0 0 2 2 0 1 0 0 5LI 0 0 0 5 1 3 3 1 0 0 13LT 1 0 0 6 2 1 3 1 0 0 14RT 0 0 1 3 1 1 17 2 0 0 25RI 0 0 1 6 2 7 9 5 0 1 31

bit

uati

on RI 0 0 1 6 2 7 9 5 0 1 31

RM 0 0 1 5 2 4 1 3 2 0 18RR 0 0 0 0 1 2 0 1 0 0 4RL 0 0 0 1 0 0 0 1 0 0 2

Total 1 0 4 27 13 22 35 15 2 1 120

s o

n H

ab Distributions of before/ after preference change

35

40

Before After

An

aly

sis

10

15

20

25

30

Freq

uenc

y

A

39

0

5

LL LR LM LI LT RT RI RM RR RL

Finger Index

Page 40: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results(BSPA)re

fere

nc

Finger Preference ChangeAll

Before LL LR LM LI LT RT RI RM RR RLLL 0 0 0 0 0 0 0 0 0 0

After preference changed

n a

nd

Pr LL 0 0 0 0 0 0 0 0 0 0

LR 0 0 0 0 0 0 0 0 0 0LM 0 0 1 0 0 0 0 1 0 0LI 0 0 0 6 0 1 0 0 0 0LT 0 0 1 0 8 1 0 0 0 0RT 0 0 1 0 2 12 7 5 2 0

bit

uati

on RI 0 0 0 0 0 1 33 5 1 0

RM 0 0 0 4 1 3 0 23 2 0RR 0 0 0 1 0 1 0 0 6 0RL 0 0 0 0 0 0 0 0 0 0

Di t ib ti f b f / ft f h

s o

n H

ab Distributions of before/ after preference change

354045

y

Before After

An

aly

sis

5101520253035

Freq

uenc

y

A

40

05

LL LR LM LI LT RT RI RM RR RLFinger Index

Page 41: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results – Habituation re

fere

nc

Foreground Ratio (F)

N f d

Number of pixels in foreground area

Denominator

Numerator

n a

nd

Pr (%)100×

×=

HWN

F foreground

Image width(=320)

Image height(=352)

Overlapped area

bit

uati

on

Sample Foregrounds of 4 Groups

0.25

Group1 Group2 Group3 Group4

s o

n H

ab

0.15

0.2

uenc

y

Enrolled sample Tested sample

An

aly

sis

0.05

0.1Freq

u

A

41

00% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Ratio of area

Page 42: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results – HabituationDenominator

refe

ren

c

Region of Overlap: Method1

lNNumber of pixels in overlapped area

Denominator

Numerator

n a

nd

Pr (%)1001

_

×=enrolledfg

overlap

NN

MNumber of pixels inforeground area of enrolled image

Overlapped area

bit

uati

on enrolled image

M1: OA/Seg_Enrolled

0.25

Group1 Group2 Group3 Group4

s o

n H

ab

0.15

0.2

uenc

y

Enrolled sample Tested sample

An

aly

sis

0.05

0.1Freq

u

A

42

00% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Score

Page 43: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results – Habituationre

fere

nc

Region of Overlap: Method2

(%)1002 ×= overlapNM

n a

nd

Pr (%)1002

__

×−+ overlaptestedfgenrolledfg NNN

M

Number of pixels in overlapped areaNumber of pixels inforeground area of tested image

Number of pixels inforeground area ofenrolled image

bit

uati

on enrolled image

M2: OA/ (Seg_Enrolled+Seg_Tested- OA)

0.25

Group1 Group2 Group3 Group4Denominator

Numerator

s o

n H

ab

0.15

0.2

uenc

y

An

aly

sis

0.05

0.1Freq

u

Overlapped area

A

43

00% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

ScoreEnrolled sample Tested sample

Page 44: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results – Habituationre

fere

nc

Region of Overlap: Method3

(%)1003 imageoverlap NNM

Number of pixels in overlapped areaNumber of pixels in whole image

n a

nd

Pr (%)1003

__

××=enrolledfg

image

enrolledfg

overlap

NNM

Number of pixels in foreground area of enrolled image

bit

uati

on

M3: M1/Foreground_Enrolled

0.25

Group1 Group2 Group3 Group4

s o

n H

ab

0.15

0.2

uenc

y

An

aly

sis

0.05

0.1Freq

u

A

44

00% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Score

Page 45: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Analysis Results – Habituationre

fere

nc

For All GroupsAnalysis results for Group1

Foreground of samples M1(OA/Enrolled)M2(OA/ (Enrolled+Tested- OA) M3(M1/Foreground Enrolled)

Analysis results for Group2

Foreground of samples M1(OA/Enrolled)M2(OA/ (Enrolled+Tested- OA) M3(M1/Foreground Enrolled)

n a

nd

Pr

0.15

0.2

0.25

eque

ncy

M2(OA/ (Enrolled Tested OA) M3(M1/Foreground Enrolled)

0.15

0.2

0.25

eque

ncy

M2(OA/ (Enrolled Tested OA) M3(M1/Foreground Enrolled)

bit

uati

on

0

0.05

0.1

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Ratio of area

Fre

0

0.05

0.1

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Ratio of area

Fre

s o

n H

ab

Analysis results for Group3

0.25

Foreground of samples M1(OA/Enrolled)M2(OA/ (Enrolled+Tested- OA) M3(M1/Foreground Enrolled)

Analysis results for Group4

0.3

Foreground of samples M1(OA/Enrolled)M2(OA/ (Enrolled+Tested- OA) M3(M1/Foreground Enrolled)

An

aly

sis

0.1

0.15

0.2

Freq

uenc

y

0.1

0.15

0.2

0.25Fr

eque

ncy

A

45

0

0.05

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Ratio of area

0

0.05

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Ratio of area

Page 46: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Observationsre

fere

nc

Little difference in distributions of personal information between Purdue and Inha U i iti t th

n a

nd

Pr Universities except the sensor useness

may affect the analysis results on habituation.

bit

uati

on

As users get experienced, they prefer right or left index fingers most

s o

n H

ab left index fingers most.

This particular sensor becomes inconvenient to

An

aly

sis This particular sensor becomes inconvenient to

use thumb fingers depending on the placement condition.A

46

condition.

Page 47: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Conclusionsre

fere

nc

Finger preferences of users for fingerprint sensor exist, and the preferences of Asians and W t it i il

n a

nd

Pr Westerns are quite similar.

bit

uati

on Exterior design of a fingerprint sensor gives an

influence to the finger preference of users.

s o

n H

ab

An

aly

sis

A

47

Page 48: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Future Worksre

fere

nc

Habituation analysis on data collected during the experiments

n a

nd

Pr Image quality distributions of individual groups

Inter-week Overlap ratio distributions of individual groups

bit

uati

on groups

ROC curves in time series for each groupSame-week Inter-group ROC curves

s o

n H

ab g p

Comparison of habituation results with

An

aly

sis p

preference results

A

48

Page 49: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Comparisonre

fere

nc

Based on transferred data BSPA INHA Comment

# of subjects (G1) 13 19 2 groups in BSPA are not satisfied protocol.# of subjects (G2) 16 20

n a

nd

Pr # of subjects (G2) 16 20

# of subjects (G3) 15 29

# of subjects (G4) 10 24

# of images/finger (G1) 15 25 BSPA did not give 7 images

bit

uati

on # of images/finger (G1) 15 25 BSPA did not give 7 images

per group, which are used for enrollment process. Moreover in case of Group1, BSPA collected image only 5 times

# of images/finger (G2) 6 13

# of images/finger (G3) 6 13

# f i /fi (G4) 6 13

s o

n H

ab collected image only 5 times.# of images/finger (G4) 6 13

Total # of images 1323 4272 About 3.23 times

An

aly

sis

A

49

Page 50: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Comparisonre

fere

nc

Based on transferred data BSPA INHA Comment

# of subjects (G1) 13 19 2 groups in BSPA are not satisfied protocol.# of subjects (G2) 16 20

n a

nd

Pr # of subjects (G2) 16 20

# of subjects (G3) 15 29

# of subjects (G4) 10 24

# of images/finger (G1) 15 25 BSPA did not give 7 images

bit

uati

on # of images/finger (G1) 15 25 BSPA did not give 7 images

per group, which are used for enrollment process. Moreover in case of Group1, BSPA collected image only 5 times

# of images/finger (G2) 6 13

# of images/finger (G3) 6 13

# of images/finger (G4) 6 13

s o

n H

ab collected image only 5 times.# of images/finger (G4) 6 13

Total # of images 1323 4272 About 3.23 times

# of ROL pairs (G1) 8190 34200 = # of images/finger x (# of images/finger - 1) x 3(fingers)# f ROL i (G2) 1440 9360

An

aly

sis images/finger 1) x 3(fingers)

x # of subjects# of ROL pairs (G2) 1440 9360

# of ROL pairs (G3) 1350 13752

# of ROL pairs (G4) 900 11232

A

50

Total # of ROL pairs 11880 68544 About 5.77 times

Page 51: (2007) Performance Analysis for Multi Sensor Fingerprint Recognition System

eClass note for the 1st term of 2005

Comparisonre

fere

nc

Based on transferred data BSPA INHA Comment

Size of heightVarious 352

Height of BSPA was fixed to 320, which is the

n a

nd

Pr minimum.

# of iterations for G15 6

Durations are the same as 6 weeks between the first and the last.

bit

uati

on the last.

# of iterations for G2~4 2 2

s o

n H

ab

An

aly

sis

A

51