IEEE Consumer Electronics Magazine - July 2016 - 73

an active glint in the pupil of the eye, to show if it is a live eye or
a false video eye. An antispoofing technique for a fingerprintbased identification system using liveness detection is presented
in [8]. A pulse oximeter sensor that interacts with the finger
while the fingerprint is being acquired for identification is used.
It determines a liveness factor R, determined from the oxygen
content of the hemoglobin and the heart rate, which is compared
to a threshold to decide the liveness of the biometric sample.
Secondary biometrics can provide an extra layer of security.
Daugman has published results of studies involving the order
of hundreds of billions of cross comparisons of iris codes [9]
showing that while one iris code might be duplicated across a
large segment of the population, there is almost no statistical
likelihood for a pair of people to have both iris patterns duplicated. This does not occur even in the case of identical twins.
A similar logic follows if we use two different biometrics,
for example, iris and palm print or iris and fingerprint. Building
a multimodal-biometric system that combines a good biometric,
such as a fingerprint, with another biometric, such as a palm
print, is one of the most interesting topics of research these days
[10], [11]. That brings us to the core hypothesis of this article.

PALM PRINTS AND SMARTPHONES
Today the smartphone industry has successfully implemented fingerprint biometrics. However, fingerprints are one of
the easier biometrics to fake-we leave copies of our fingerprints all around us on a daily basis as we touch and handle objects. There have been various high-profile examples
of different ways to trick fingerprint sensors, and there is a
general consensus that while fingerprints are very convenient and their use case is now well developed for smartphones, it would not be acceptable to rely on fingerprints
alone for high-value transactions such as personal banking
or financial transactions.
The research into second-generation biometrics is primarily focused on the iris of the eye [12]-[14]. However, there is a
need for a secondary biometric that can be combined with the
iris for additional authentication. One major barrier to adding
additional biometrics is that each requires a new sensing technology. Fingerprints required that suitable sensing technology
was available, as the initial experience with swipe sensors was
highly unsatisfactory, and iris acquisition will require either
modified optical designs or the use of new NIR-sensitive
CMOS image sensors, or possibly a combination of both.
New hardware designs are expensive and take time to perfect, delaying the market entry of a new biometric technology. So what if there were a biometric that didn't need new
hardware? What if there were a biometric that doesn't leave a
trail of exemplar on everything we touch?
That brings us to the thinking behind this article-our
palm prints satisfy the above criteria, and while palm print has
not seen the mainstream use of face, fingerprint, and iris biometrics, it is an equally valid candidate. And what could bring
it into prominence is that it requires no modifications to a conventional smartphone-the main camera on any of today's
phones is ideally suited to acquire a high-quality image of

your palm print with more than sufficient resolution to determine the main features of your palm. We won't be able to tell
your future, but we should be able to authenticate you.
In this article, a proof-of-concept use of palm-print-based
authentication for smartphones is presented. This is a preliminary study, but tests are across a handful of different devices
and the core hypothesis is that palm print recognition can be
device independent and that any of today's devices can be used
to validate a user.

PALM PRINT AS A BIOMETRIC
As discussed in the previous section, palm print can, in theory,
replace or complement the more common biometrics like fingerprint or iris for authentication in mobile devices. Palm prints
have plenty of features that can be used in the recognition process, like texture, principal lines, secondary lines (wrinkles),
and ridges. The palm area is large enough to allow for features
extraction even from images acquired with low-resolution cameras. Plus, it doesn't require extra hardware. The rear camera of
any smartphone should work.

LITERATURE REVIEW
Although not as popular as fingerprint or iris, palm print as a
biometric has generated interest within the research community, and some interesting methods have been reported in this
area. For those who are interested, we recommend two palm
print literature surveys, [15], [16], with the observation that
these do not cover smartphone-based approaches.
The initial research was mostly done on publicly available
palm print databases taken under very controlled conditions,
where the extraction of region of interest (ROI) is not an issue.
Several methods for feature extraction, for example, competitive code (CompCode) [17], ordinal code [18], and derivative of
Gaussians [19], were applied to the Hong Kong Polytechnic
University (PolyU) Database [20] with very good results-
equal error rate (EER) of less than 0.5%. Other algorithms have
been successfully implemented in the Indian Institute of Technology Delhi Database [21], like local binary pattern (LBP)
[22], SIFT features [23], and a combination of SIFT and
orthogonal line ordinal features (OLOF) [24]. The LBP-based
approach reaches an EER of around 3.8%, while the SIFT features have an EER of around 1%. The combination of OLOF
and SIFT further improves the EER to 0.18%.
In recent years, the smartphone has been considered as a
means of acquiring palm print images. In [25], the authors have
created their own databases using a mobile phone. To normalize
the hand's posture variation, a hand-shaped guide window was
used on the screen, and the embedded flashlight was turned on
during the image acquisition. The ROI features have been
extracted using local orientation histograms of palm lines. They
reached an EER of 2.88%, while CompCode and OLOF provide EERs of 6% and 5%, respectively. A square guide window
was employed in [26] for image acquisition, and band-limited
phase-only correlation was used for feature extraction. The
authors compared the efficiency of their algorithm on their database against PolyU and CASIA [27], and the reported results
JULY 2016

IEEE Consumer Electronics Magazine

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - July 2016