International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp. 11279-11284
© Research India Publications. http://www.ripublication.com
11279
Performance Evaluation of PPG based multimodal biometric system using
modified Min-Max Normalization.
Girish Rao Salanke N S
1
, Dr. M V Vijaya Kumar
2
, Dr. Andrews Samraj
3
1
Assistant Professor, Department of Computer Science & Engineering, R V College of Engineering, Bangalore, India.
2
Professor, Department of Computer Science & Engineering, Dr Ambedkar Institute of Technology, Bangalore, India.
3
Director, Advance Science and Technology Research Center, Mahendra College of Engineering, Salem, India.
Abstract
Usage of Photoplethysmograhy (PPG) signal which was
limited for clinical purposes is explored for the biometric field
by fusing it with a traditional biometric such as fingerprint. A
multimodal biometric system is proposed to overcome the
limitations of unimodal biometric system. A modified Min-
Max Normalization score level fusion is proposed for
multimodal biometric system. The paper evaluates the
performance of PPG based multimodal biometric approach
where in it is observed that the False Acceptance Rate of
fingerprint biometric system is reduced from 5.4 % to 3 % and
similarly the False Rejection Rate is reduced from 6.7 % to
3.8 % by fusing the PPG component with fingerprint. The
proposed method exhibits good identification accuracies when
PPG signal is used as one of the biometric trait in a
multimodal biometric system.
Keyword: Biometric, PPG signal, Score Fusion, False
Acceptance Rate (FAR), False Rejection Rate (FRR), Min-
Max Normalization
INTRODUCTION
The traits used in multimodal biometrics[1] have relayed more
on traditional biometrics like fusing physiological features
such as face with fingerprint, face with palmprint[2], and face
with Iris and so on. Most researches in the biometric
community have ignored the intrinsic characteristics of the
biological signal for their applications. Studies of such signals
that can be used for biometrics are very important. Some of
the signals that can be considered are ECG, EEG and PPG[3]
signals, which exhibit a rich set of features that can be used
for identification and verification purpose. The second
objective of this work is to propose a new algorithm that is
robust to day today as PPG changes due to motion artifact.
Since the PPG signal is a time series there is always the
question of how long should the PPG is acquired. Considering
that enrolment is done only once, subjects will agree to spend
some time enrolling themselves into the system however for
verification our goal is to minimize the authentication time for
the subjects. The final objective of this work is to explore the
effectiveness of using the PPG signal in multimodal biometric
systems. Since there is no biometric which has 0% False
Rejection Rate and all biometrics have their own limitations
and disadvantages, multimodal biometric systems uses more
than one biometric traits for the sake of improve the
performance and making the system robust to spoof attacks.
The main challenge is that to choose appropriate biometrics
such that the inherent weaknesses can be offset by overall
system design. Therefore in order to further improve the PPG
biometric system we propose a multimodal biometric system
by fusing PPG and fingerprint. The fingerprint matcher offers
high accuracy in terms of authentication however suffers from
spoof attacks since a fingerprint trace can be easily taken from
any surface that a finger has touched. Finally both modalities
can be collected conveniently from subject’s fingertips which
require less cooperation from subjects unlike other systems.
Multimodal biometrics[4] combines information from
different sources as compared to unimodal[5] wherein person
recognition is based on a single source of biometric
information. Some of the system requirements are not meet in
Unimodal biometrics; therefore combining multiple
biometrics can overcome the limitations of unimodal
biometrics and also improve the performance of the overall
system. In multimodal biometrics the sources of information
can be from multiple sensors, multiple traits, multiple
instances or multiple instances. In multiple sensors the
different sensors are used for capturing single biometric trait.
For example face images of an individual can be captured
using two different sensors. In multiple traits, the system
information from different biometric traits are combined to
authenticate a subject, for example combing face and
fingerprint. In multiple Instances, the systems use multiple
instances of a single biometric trait, such as the image of the
left and right eye of a subject for a retina recognition system.
In multiple Sample, the system uses a single sensor is used to
capture multiple samples of a single biometric characteristic
of a subject, for example frontal, left and right profiles used in
face recognition
The fusion[6] in multimodal biometrics can be done at
different levels, such as
a) Feature level fusion
b) Decision level
c) Score level.
In Feature level fusion feature set extracted from multiple data
sources are combined to create a new feature set as shown in
figure 1. If the features from different biometrics traits are in
the same type of measurement than it is recommendable to
combine there features vectors into a single new vector.