|
|
DaonAnalytics
Although trials and deployments provide large volumes of raw biometric data,
considerable offline processing and analysis is necessary in order to calculate
and understand biometric performance and accuracy in different working environments.
Currently there are no independent toolkits to carry out this evaluation
therefore in order to automate this huge analysis task Daon has created
DaonAnalytics.
DaonAnalytics is a software product designed to support offline biometric
statistical analysis activities based on logged biometric data. The product
allows customers such as prime integrators to automate the production of
Receiver Operating Characteristic (ROC) and Cumulative Match Characteristic
(CMC) curves.
-
A Receiver Operating Characteristic (ROC) curve plots the probability
of a false-match against that of a false non-match for a single match
attempt. The ROC shows the range of false match rate (FMR)/false non-match
rate (FNMR) tradeoffs available for a given matching algorithm on images
captured from a specified device, in a given environment and population.
-
A Cumulative Match Characteristic (CMC) curve plots the probability of
identification against the returned 1:N candidate list size. It shows
the probability that a given user appears in different sized candidate
lists. The faster the CMC curve approaches 1, indicating that the user
always appears in the candidate list of specified size, the better the
matching algorithm.
DaonAnalytics Functionality
The diagram below shows the high-level components and information flow
as image samples are analysed by the tools to produce accuracy curves
and other statistics. The four main steps involved are:
-
Collection of
biometric data and the description of that data. The Daon Client can log
biometric data in the required format, or other applications may be used
to capture such data
-
Filter the biometric data, by removing certain
data depending on specific dates, times, usage, users, and quality scores
etc. Filtering allows "what if" style analysis, such as comparing results
with thumbs against results with right index fingers, or performance of
certain populations or age groups against others. The implementation of
filters allows a customer to effectively model a variety of deployment
protocols (e.g how many attempts, what threshold, population subsets etc)
in the most cost-efficient and accurate way.
-
Generate the templates
corresponding to the biometric data. One template set (e.g. enrolment
templates) is then matched against another (e.g. verification templates)
using a specific matching algorithm. All the matching scores are saved
for analysis purposes in the next step
-
Analyse the matching scores to
calculate biometric accuracy and error rates, and plot this performance
on a ROC or CMC graph
If you would like to receive further information on DaonAnalytics please
.
|
|
|
|
|
|
|
|
|
|
|
