Ear biometrics pattern

recognition identification
system using Fuzzy Logic
Ariel Carrasco
Ray Gonzalez
Intro: Pattern recognition
Pattern: physical or abstract repeated structure of an object. It
can be distinguished from others by their features.

Features: Distinctive attributes.

Pattern recognition Process of classifying or identifying input

data, assigning them into classes based on features.
Pattern recognition block
diagram
Training data, Establish the association between
features and classes.
Test data which is used observe the overall
performance of the system.
 Fuzzy logic for pattern
recognition
Classical models of these type of systems
identify a pattern into a class depending on the
features similarity.
Pattern classes can overlap causing problems in
object recognition.
Fuzzy Logic pattern recognition can
classify/identify patterns into one or more
classes with a certain degree of membership to
belong in each class.
Biometric systems
Biometrics is the measurement of
people unique physical and behavioral
characteristics used to recognize
individuals identity for security or
access control applications.

Ideal biometrics characteristics are the

following:
Universal
Unique
Permanent
Collectable
 Problem
Every biometric system has his drawbacks..

Biometric Disadvantages
systems
DNA Not real time
identification.
Fingerprint Passive
identification
(Requires
placing finger
on a
device).Childr
en fingerprint
size changes
quickly
Face Active
recognition identification,f
ace features
 Shape of the ear focusing on details like the helix
Ear biometrics rim, lobule, antihelix, concha, etc., geometric
measurements or certain sections of it.
Ears are unique, they become fully formed at a
young age 8 years old) and doesnt change over
time (lobe descends a little, it grows again after
age 70 ).
Computer vision
No interaction with the device

Drawbacks
Camera location, appropriate lightning, high
resolution.
Hair, earings.
Significance of the project
Biometric authentication offers many
advantages over current security
practices like identification cards, keys,
tokens, passwords,PIN, etc.

Biometrics are useful in business

authentication needs, to substitute
undocumented access, ID swapping,
credential replacements, manual badge
checks, etc.
 Fuzzy Inference System Aided Perso
Methodology nal Detection: Ear Biometrics

16Mega pixel camera, at 300

Pixels per inch was used
mm=(number of pixel*25.4)/PPI
Database: 840 images from ear
the ear.
1m fixed position.
5 inputs: Major, minor axis, ratio
of upper area versus lower area
range. radius gyration of x axis
and y axis.
Test sample image :20
420 Fuzzy inference rules.
Mamdani
Used to classify ears into 4 main
groups
Image processing
1. Convert color image to grayscale
image.
2. Detect the edge (Canny edge
detection), inverting the image then
obtain clear edge of ear image.
3. The extracted features of fuzzy
linguistic values are stored in the
database.
Edge detection algorithm process:
1. Gaussian filter,average filter (Smooth
the image to remove noise)
2. Find intensity gradients of the image
(detection of horizontal vertical and
diagonal edges of image).
3. Apply thresholds for border
enhancement.
 Geometric features
Major and minor axis, ratio of upper area versus
lower area range. radius gyration of x axis and y
axis.
Clustering K-means
The purpose of applying k-means clustering
algorithm is to find a set of clustered centers
and a partition of training data into
subclasses. So how k-means basically works
is that k initial are randomly generated
within the data domain. K clusters are
created by associating every observation
with the nearest mean. The centroid of each
of the k clusters become the new mean.
 Fuzzy system
model
Major, minor axis, ratio of upper area versus
lower area range. radius gyration of x axis
and y axis.
Fuzzy system rules (if-Then)
Fuzzy system output
 Results

Sensitivity: measures the proportion of positives that are correctly identified as such.(the percentage of ears
that are correctly identified as belonging into a group).

Specificity:measures the proportion of negatives that are correctly identified as such.(the percentage of ears
that are correctly identified as not belonging into a group).

Accuracy. proportion of true results (true positives and true negatives) among the total number of cases .
Discussion/Conclusions
Personalized fuzzy logic ear recognition requires a very specific membership
function output to obtain a high degree of correlation. between the new ear (input)
and database ear.

Solution:

More feature extractions of the ear, (More fuzzy rules) to improve sensitivity,specificity and
accuracy

Helix: Length at certain

points
Discussion/Conclusions

Multisensor ear pattern recognition, using image processing and acoustic sensor, (More
fuzzy rules) to improve sensitivity, specificity and accuracy

Sends an acoustic wave which

will bounce off the tympanic
membrane and are read by
microphone..

Drawback: Computing time.

Improvements on the paper: Automatic ear detection and cropping.

References
[1]T. Ross, Fuzzy logic with engineering applications, 3rd ed. Chichester, West
Sussex: Wiley, 2010.

[2] Pedrycz, W. "Fuzzy Logic In Development Of Fundamentals Of Pattern

Recognition". International Journal of Approximate Reasoning 5.3 (1991): 251-264.
Web.

[3] M.A.Jayaram,Prashanth G.K Cluster Based Ear Biometric System for Personal
Identification http://www.jmest.org/wp-content/uploads/JMESTN42351157.pdf. Web

[4] Prashanth G.K, M.A. Jayaram, Rohit K Fuzzy Inference System Aided Personal
Detection: Ear Biometrics ISSN: 2347-5552, Volume-4, Issue-5, September-2016.
Web