(2013) trade-off between impression numbers and attempt numbers

Post on 25-Jun-2015

86 Views

Category:

Technology

1 Downloads

Preview:

Click to see full reader

DESCRIPTION

Presented at ICITA 2013, 8th International Conference on Information Technology and Applications, Sydney, Australia, 1 - 4 July, 2013 The amount of time taken to enroll or collect data from a subject in a fingerprint recognition system is of paramount importance. Time taken directly affects cost. A trade-off between number of impressions collected and number of interaction attempts allowed to submit those impressions must be realized. In this experiment, data were collected using an optical fingerprint sensor. Each subject submitted six successful impressions with a maximum of 18 interaction attempts. The resulting images were analyzed using three methods: the number of interaction attempts per finger, quality differences from the first three impressions to the last three impressions, and finally matching performance from the first three impressions to the last three impressions. The right middle finger seemed to have the most issues collecting as it required the most interaction attempts. Analysis was performed to show no significant differences in image quality or matching performance. However, after further analysis, a steady improvement was noticed from Group A to Group B in both image quality and matching performance.

TRANSCRIPT

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

07/02/2013Jacob A. Hasselgren, Stephen J. Elliott, Jue Gue

TRADEOFF BETWEEN IMPRESSION NUMBERS AND ATTEMPT NUMBERS

AGENDA

• Introduction• Methodology• Results• Lessons learned• Next steps

INTRODUCTION

• Many factors can impact the performance of a biometric system from poor quality data [1]:– Skin conditions [2]– HBSI [3]– Associated meta-data [4]

• Collect appropriate data and minimize time/error

MOTIVATION

• The efficiency of a biometric system is important

• While investigating habituation of biometric systems, the issue of the proper number of impressions collected and time arose

MOTIVATION

• We collect everyday – at what point do you stop collecting from a subject?– 3 samples?– 6 samples?– 9 samples?

• We cannot keep subjects forever due to time and costs

MOTIVATION

• Number of subjects is important• Must design protocol to keep costs low

while still processing as many subjects as possible

• How many impressions should be collected from each subject?

• How many attempts should be allowed?• Which fingers should be collected?

QUESTIONS

• Are some fingers harder to collect? – Do some take longer?

• Is the image quality of the first three any different from the last three?

• Does the performance change when slicing groups taken from subjects?– ie. Do the first three impressions match well

against the last three impressions of the same subject?

DATA

• Collection of multimodal samples

• Only data from one fingerprint sensor is be used for analysis

• U.are.U 4500

• Taken from ongoing aging study in the BSPA Labs

DEFINITIONS

Term Meaning

SAS Successfully acquired sample, synonymous with impression

Impression number Number given to each SAS, ranging from 1-6

Interaction attempt number Number given to each interaction attempt, whether it results in success

or failure, ranging from 1-18

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

METHODOLOGY

METHODOLOGY

• 6 impressions/SAS were taken on each finger with a maximum of 18 interaction attempts to do so

• Fingers used:– Left index– Left middle– Right index– Right middle

METHODOLOGY

• An impression number was given to each SAS in order for each finger– Six impressions per finger, impression

numbers range 1-6

• An interaction attempt number was given to each interaction attempt in order for each finger– Max of 18 interaction attempts was given to

submit 6 SAS’s, intearction attempt numbers range 1-18

METHODOLOGY

• A number of approaches were used to analyze the samples– An analysis of the number of attempts

required for each finger– Compare quality scores between designated

groups (first three, last three)– Compare matching rates between same

designated groups

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

RESULTS

NUMBER OF INTERACTION ATTEMPTS

• A comparison of the number of SAS’s to the number of interaction attempts was performed

• Each subjects submitted six SAS’s• Did any given finger require more

interaction attempts than others?

NUMBER OF ATTEMPTS

NUMBER OF ATTEMPTS

• No significant difference between the fingers was found

• The right middle finger seems to have the most interaction attempts of all of the fingers collected– As well as the most variance

DISTRIBUTION OF QUALITY

• A comparison of the first three collected samples for any given finger to the last three was performed to find differences in image quality

DISTRIBUTION OF QUALITY

• All of the samples were pushed through a quality scoring algorithm, Aware WSQ 1000

• Scores a number of different metrics, with an overall quality score

• This overall quality score was used in the following analysis

DISTRIBUTION OF QUALITY

DISTRIBUTION OF QUALITY

• A one-way ANOVA was used to compare the first three SAS to the last three

• No finger showed a significant difference

Finger P-Value

RI 0.155

RM 0.460

LI 0.090

LM 0.050

MATCHING PERFORMANCE

• A comparison of the first three collected samples for any given finger to the last three was performed to find differences in matching performance

MATCHING PERFORMANCE

• The samples were enrolled into a minutiae-based matching software, Megamatcher 4.3

• The first three SAS’s and last three SAS’s were enrolled separately

• Equal error rates were used for results

MATCHING PERFORMANCE

Finger First 3 vs First 3 Last 3 vs. Last 3 First 3 vs. Last 3

LI 0.0000 0.0000 0.0006

LM 0.3322 0.0000 0.1282

RI 0.0000 0.0000 0.0000

RM 0.0000 0.0000 0.0000

MATCHING PERFORMANCE

• No improvements were noticed in performance for any fingers except for left middle– The EER improves from .3322 to .0000

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

LESSONS LEARNED

LESSONS LEARNED

• No issues were found when comparing the number of interactions attempts between fingers– Though, right middle finger may require

more attempts to collect impressions

• To save time and unnecessary costs, it may not be practical to collect additional samples– No significant improvement was found after

collecting three additional samples

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

NEXT STEPS

NEXT STEPS

• Replicate this study on more sensors• Attempt to observe habituation using

more than one visit– Observe effect of habituation on attempt

numbers

• Include time-on-task to get a true estimation on time required to collect SAS

• Include hand dominance in interaction attempts

BIOMETRICS LABBiometric Standards, Performance and Assurance LaboratoryDepartment of Technology, Leadership and Innovation

QUESTIONS?

jahassel@purdue.edu

top related