Biometric system based on one single large area a-SiC:H p-i-n photodiode

chainbirdinhandSécurité

23 févr. 2014 (il y a 3 années et 6 mois)

82 vue(s)

Biometric system based on one single large area a
-
SiC:H

p
-
i
-
n photodiode


M. Vieira


mv@isel.pt

M. Fernandes


mfernandes@isel.pt

A. Fantoni


af@isel.pt


P. Louro


plouro@isel.pt

Electronic Telecommunication and Computer Dept.
ISEL, Rua Conselheiro Emídio

Navarro, P 1949
-
014 Lisboa,
Portugal



Abstract

Based on the Laser Scanned Photodiode (LSP) image
sensor we present an optical fingerprint reader for
biometric authentication. The device configuration and the
scanning system are optimized for this specif
ic purpose.

The scanning technique for fingerprint acquisition is
improved and the effects of the probe beam size,
wavelength and flux, the scan frequency on image contrast
and resolution will be analyzed under different electrical
bias. An optical model
of the image acquisition process is
presented and supported by a two dimensional simulation.

Results show that a trade
-
off between read
-
out parameters
(fingerprint scanner) and the biometric sensing element
structure (p
-
i
-
n structure) are needed to minimiz
e the
cross talk between the fingerprint ridges and the
fingerprint valleys. In the heterostructures with wide band
gap/low conductivity doped layers the user
-
specific
information is detected with a good contrast while the
resolution of the sensor is aroun
d
20

m
. A further
increase in the contrast is achieved by slightly reverse
biasing the sensor with a sensitivity of
6.5

Wcm
-
2

and a
flux range of two orders of magnitude.


Keywords

Non
-
pixelled image sensors, Laser Scanned Photodiode
image sensor, biometric s
ystem.



INTRODUCTION

Studies on the use of glass/ITO/p

/i

/n

/metal

structures as
Laser Scanned Photodiode (LSP) image sensors have
recently shown its potential capability [
1
,
2
] as
monochrome image sensing devices. These devices are
fundamentally differe
nt from the other electrically scanned
image sensors [
3
,
4
]
. They consist on one single large cell
detector or pixel (picture element) and the image is scanned
by sequentially detecting scene information at discrete XY
coordinates. The read
-
out of the inje
cted carriers is
achieved by measuring the ac component of the current,
i
ac
,
in short circuit mode. They detect an optical image
(monochrome image) with a spatial resolution of 30

m in
a flux range of two orders of magnitude and present a


sensitivity of

6.5

Wcm
-
2
. Advantages to this approach are
high resolution, uniformity of measurement along the
sensor and the cost/simplicity of the detector. The design
allows a continuous sensor without the need for pixel
-
level
patterning, and so can take advantage o
f the amorphous
silicon technology. It can also be integrated vertically, i. e.
on top of a read
-
out electronic, which facilitates low cost
large area detection systems where the signal processing
can be performed by an ASIC chip underneath.


Figure 1 depi
cts the sensor configuration, in addition the
electrical model proposed and the depletion regions (dotted
lines) are visualized.


I
sc
n
i

p
i
ph
-
Scanner
Image
1/
r
d
~
I
ph
R
Ln
Al
R
Lp
D
Depletion
region

Figure 1.

Laser Scanned Photodiode configuration and
electrical model.


This work aims to clarify

possible improvements, physical
limits and performance of the LSP image sensor when used
in a biometric system for fingerprint authentication. Here,
the biometric capture device and the scanning reader are
optimized and the effects of the sensor structure

on the
output characteristics discussed.



THEORY OF OPERATION
AND BIOMETRIC
REPRESENTATION

The LSP operation and image representation are based on
the analysis of the electrical field profile induced across
the capture device by a steady
-
state light patt
ern
illumination. Low local electric fields are ascribed to
illuminated regions and high electric fields to dark zones
[
5
]. In the dark regions the carriers generated by the
scanner are separated by the electric field and collected
(high
ac

component of th
e photocurrent), while those
generated inside the illuminated regions mostly recombine
inside the bulk (low
ac

component of the photocurrent).
So, by mapping the ac component of the photocurrent,
i
ac
,

during the scanning of the capture device it is possibl
e to
reconstruct the projected biometric feature.

To understand the transport mechanism under non
-
uniform
illumination we took into account the experimental
characterization of the devices and some results from a
computer simulation performed using the Am
orphous
Solar Cell Analysis simulator ASCA

6

. 䑥瑡楬i 慢o琠瑨攠
progr慭 and 瑨攠 exp敲imen瑡氠 inp琠 p慲ame瑥ts 慲e
d敳捲楢敤 敬獥wh敲攠
[
7
]

䙩Fr攠2 d楳p污ls 瑨攠sim污瑥l
po瑥t瑩慬tprof楬攠w楴hin
慮 a
-
S椺䠯i
-
卩S䠯H
-
S楃䠠s瑲捴re
nd敲 愠 汩lh琠 spo琠 (650 nm, 1

Wcm
-
2
) illumination in
superposition to a two strip light image (650nm, 10

Wcm
-
2
).


0.0
0.1
0.2
0.3
0.4
0.5
0.0
0.1
0.2
0.3
0.4
0.5
2
4
6
8
10
12
14
Potential V
Position (

m)
Lateral position (a. u.)
V=0 V

Light image

p

n

i




Figure 2.

Simulated potential profile within a

a
-
SiC:H/a
-
Si:H/a
-
SiC:H structure under a light spot
i
llumination in superposition with a light image.


The internal configuration of this structure is different
from the standard p
-
i
-
n a
-
Si:H solar cell, since local
additional fields located at the interfaces cause
accumulation of charges, which limits the c
ollection of the
photocurrent at the illuminated regions and improves the
lateral currents within the i
-
layer. Simulated results show
that
the transport mechanism in dark conditions depends
almost exclusively on field
-
aided drift while under
illumination i
t will depend mainly on the diffusion of
carriers towards the contacts.
When the laser spot is in
superposition with the light image (assumed to be much
stronger) the electrical distribution inside the device is not
affected since the effect of the light s
pot is negligible. The
device is blind to it, producing a theoretical image read
-
out with a definition of about 10

m.



EXPERIMENTAL DETAILS

The analysed p
-
i
-
n capture devices were deposited by
Plasma Enhanced Chemical Vapour Deposition [
8
] on
glass (Corning 7059) covered with transparent aluminium
doped ZnO as front contact. After the deposition of the
amorphous laye
rs, a back contact of aluminium was
thermally evaporated which defines the active area of the
sensor (4

4

cm
2
). The deposition conditions where kept
constant for all i
-
layers, while they varied in the doped
layers by adding methane and keeping the doping l
evel at
low values. All the layers of sensor #1 are based on
amorphous hydrogenated silicon, the n
-
layer of sensor #2
and the p
-

and n
-
layers of sensor #3 are based on a
-
SiC:H
alloys. The optical gap sequences for each p
-
i
-
n sensor are
respectively 1.8/1.8
/1.8 eV, 1.8/1.8/2.1 eV and 2.1/1.8/2.1
eV. All the structures were characterized by spectral
response, current
-
voltage and capacitance
-
voltage
measurements, in dark and under different optical bias
conditions (0<

L
<2 mWcm
-
2
), as described elsewhere
[
9
]
.

A

laser light illuminates a fingerprint placed on a glass
surface (platen). The reflectance of this light is projected
onto the active surface of a p
-
i
-
n capture device and the
fingerprint image is captured directly using the Laser
Scanner Photodiode as a f
ingerprint scanner. The amount of
reflected light is dependent upon the depth of the ridges and
valleys on the glass. The light that passes through the glass
into the valleys is not reflected towards the sensor active
surface (dark regions), whereas light
that is incident upon
ridges on the surface of the glass is reflected (light regions).
The read
-
out of the injected carriers is achieved by
measuring the
ac

component of the current,
i
ac
.

The scanning and acquisition processes are controlled by a
microcom
puter which stores the current as a two
dimensional array of discrete values,
i
ac
(m,n)
, each one
representing the photocurrent induced by the chopped light
at the selected position. The image intensity,


is obtained
by subtracting the background (without image) from the
input matrix. All the measurements were performed with a
chopper frequency less than 1 kHz and a scanner
wavelength of 633 nm. The scanner
-
induced photocurrent
was measured using a lock
-
in amplifier.



RESULTS AND DISCUSSI
ON


Sensing element configuration

In order to choose the best capture device structure (homo
or hetero) for the fingerprint read
-
out, the active area of
each sensing element was homogeneously illuminated by
filtered ligh
t (

=535nm) with different intensities
(0<

L
<120

W/cm
2
), and the photocurrent generated by an
additional chopped light (

=633 nm) was measured. No
electrical bias was applied (V=0) and the read
-
out
parameters (scanner intensity, frequency and spot diameter)
where kept constant for all the sensors.

0
20
40
60
80
100
120
0.0
0.2
0.4
0.6
0.8
1.0

0
=22

Wcm
-2

0
=4.3

Wcm
-2

0
=2.2 mWcm
-2


#1 (Si:H/Si:H/Si:H)
#2 (Si.H/Si:H/SiC:H)
#3 (SiC:H/Si:H/SiC:H)
i
ac
(

L
)/i
ac
(0)

L
(

W/cm
2
)

Figure 3.

iac(

L
the image brightness,

L
.

0

is the flux constant for each
exponential fit (dash line).



Figure 3 displays the light
-
to
-
dark sensitiv
ity [i
ac
(

L
)/i
ac
(

)]
dependence on the light flux. The dash lines are exponential
fits to the data and

0

the correspondent flux constant.

In both heterostructures (#2 and #3) the signal exponentially
decays with

L

while in the homostructure (#1) no
sign
ificant decrease is detected in the flux range analyzed.

The heterostructures can detect small light
-
to
-
dark
variations while the homostructure remains “blind” in a
large flux range.

0

determines the light

to
-
dark
responsivity and can be correlated with t
he image contrast.
Results show that in the low flux range the images acquired
using sensor #3 present higher contrast than the ones obtain
by sensor #2. This makes sensor #3 suitable for biometric
applications since a good image contrast is required at lo
w
flux ranges.


Bias and image brightness voltage dependences

In order to analyse the behaviour of the sensor under
different electrical bias the active area was homogeneously
illuminated by filtered light (

=630nm) with different
intensities

L

and the photocurrent generated by an
additional chopped light (

=630nm) was measured.


As expected, the ac component of the photocurrent
monotonically decreases with the increase in the bias
voltage independently of
the light intensity. However, if we
look at the signal difference between dark and illuminated
conditions (Figure 4), the maximum contrast (signal
difference) depends on the bias voltages and is achieved
with slightly negative bias for high values of

L

an
d
positive bias for low values of

L
.


0
2x10
-3
4x10
-3
6x10
-3
8x10
-3
10x10
-3
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
3
6
9
12
i
ac
(0)-i
ac
(

L
) (a.u.)
Voltage (V)

L
(

W cm
-2
)


Figure 4
.

i
ac
-
i
ac
(

L
)
dependence on electrical bias and
image brightness,

L
.


Figure 5 shows the image intensity as a function of the
image brightness under different bias

voltages. The solid
lines are exponential fits to the data. Results show that the
image intensity presents a good linearity (linear regime) at
low flux levels and starts to saturate around

0

(saturation
regime)
.

The turning point between the two regimes
increase as the applied bias goes from reverse to forward.


0
2
4
6
8
10
12
14
16
18
0.0
0.3
0.5
0.8
1.0
1.3

0
#3 (SiC:H/Si:H/SiC:H)

=535 nm

V=-0.3 V
V=0 V
V=+0,3 V

(a.u.)

L
(

W/cm
2
)

Figure 5.

Relative Image intensity as a function of

the
image brightness,

L



The image intensity depends on the applied bias and
incr
eases with the increase of the brightness of the projected
images.

amplitude has a maximum,

max
, near short
circuit conditions and decreases with the increase of both
the reverse and forward applied bias. It is interesting to
notice that the maximum pos
ition linearly shifts from
reverse to forward voltages as

L

decreases.

By applying electrical bias a trade
-
off between linearity and
image responsivity may be established since, in the linear
regime, the reverse bias increases the amplitude of the
saturation

image a
nd the forward bias enhances its contrast. At high

L
, the signal saturates easily due to the maximum
shrinkage of the depletion region (flat band condition). So,
if the sensor is biased in reverse mode,

0

increases and the
flux range before the saturatio
n regime is enhanced
increasing the sensor linearity.


Figure 6.

Measured ac photocurrent for one
-
dimensional
scans under forward, reverse and zero bias.


At low values of

L

it is difficult to reach the saturation
regime. By s
lightly forward biasing the sensor the
collection in the illuminated regions is decreased while in
the dark ones it remains almost the same leading to a best
light
-
to
-
dark responsivity. In both cases

will be enhanced
(Figure 4).

In Figure 6 are represent
ed the scans of a pattern composed
by dark and illuminated regions projected on the active area
of the sensor under forward, reverse and zero bias voltage.
A compromise between the brightness of the projected
image and the applied bias is needed to enhance

the image
representation. For biometric applications (very low flux
range) the sensor performance (linearity and contrast) could
be improved with increased signal amplitude under slightly
reverse bias.


Line scanning frequency

When a fast scanning speed i
s necessary, the lock
-
in
amplifier cannot be used because of the limitation in the
modulation frequency. In this case a low noise current to
voltage converter is used and the separation between the
currents due to the image and to the scanner beam is
perfo
rmed after by a simple subtraction. With this method,
the overall intensity of the image as to be controlled in
order to avoid the saturation of the amplifier.

Figure 7 shows the result of a single line scan in dark (line),
and crossing an illuminated are
a (squares). The signal
represents the photocurrent due to only to the scanner; the
offset, which exists due to the steady state illumination, was
subtracted. The current was measured across a 1 k


resistor, which limits the rise and fall times measured to

about 100

s.

0,0
0,5
1,0
1,5
2,0
0,0
0,2
0,4
0,6
0,8
1,0
Light
Sensor


Signal (a.u.)
Time (ms)

Figure 7
. Single line scans in dark (line) and crossing an
illuminated region (squares).


In this case the line scan frequency is close to 1 kHz with a
small reduction in the resolution caused by the value of the

rise and fall times, which can be enhanced by the reduction
of the load resistor.

If one considers a 100 lines image then a frame rate of
10

Hz is obtained which is good enough in some
applications.


Acquisition time and scanner diameter


Figure 8 present
s the normalized signal obtained for a
one
-
dimensional scan on a sensor illuminated with a 5
-
mm
strip using two different scanning beam diameters.

0
2
4
6
8
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Dark
Region
Illuminated
Region


Spot size

1 mm
0,05mm
Normalized Signal
Position (mm)

Figure 8.

Normalized signal for one
-
dimensional scan with
two scanner beam sizes.


As one can see the increase in the diameter leads to a less
steep variation of the signal when the scanner crosses the
transition region between the illuminated and the dark zone.
This fact leads to a blurring effect when the image is
obtained as a gray
level picture, but has no effect for black
and white pictures if the threshold level is chosen at half the
signal variation.

It was shown [
10
] that the time needed to acquire an image
(T) increases with the area to scan (A), with the
bandwidth

frequency (

f
) and decreases with the scanner diameter
(T

A
.

f

d
-
2
).



BIOMETRIC REPRESENTA
TION

So a compromise between image quality and acquisition
time has to be taken into account for a specific application.
Figure 8 shows a fingerprint representation obtained
with
sensor #3 under

0.3 V, 0V and 0.3 V applied electrical
bias. The maximum light flux in the fingerprint image was
of the order of 5

Wcm
-
2
. No image processing algorithms
were used.


Figure 7.

Image representation of a fraction of a
fingerprint ac
quired with sensor #3 under reverse (V=
-
0.3
V), short circuit (V=0 V) and forward (V=0.3 V) modes.


As expected the best fingerprint representation was
obtained under short circuit conditions were the trade off
between the image brightness (10.7

Wcm
-
2
) an
d the
applied bias (V=0) is established by the

max

position (see
Figure 4).



CONCLUSIONS

Based on the LSP we have presented an optical fingerprint
reader for biometric authentication, where the capture
device configuration and the scanning reader were
o
ptimized for this specific purpose.

Data reveals that the performance of the capture device is
enhanced by a tight control of image brightness and applied
electrical bias. A trade
-
off between flux range and image
responsivity is required for a correct ima
ge reconstruction.



FUTURE WORK

Basic image processing algorithms should be applied for
image enhancement and pattern recognition. Modules for
brightness calibration and edge enhancement are still
needed to improve the biometric system. Further
optimiza
tion of the optical scanning system includes
dynamic characterization of the sensor (readout frequency,
frame rate). A new sensor structure and a prototype for
manufacturing should also be developed in order to improve
an higher resolution and

full c
olor
detection.

In Figure 8 it is represented the image of the acronym of the
sensor and a photo representation collected under short
circuit conditions, showing that it is possible to combine
several features such as signature, photo and fingerprint
representa
tions for biometric recognition. The scanning
methodology allows also to recognise in real time the
hand
-
writing peculiarities.




Figure 8.

I
mage of the acronym of the sensor and a
photograph representation collected under short c
ircuit
conditions.



ACKNOWLEDGEMENTS

We would like to thank IPE for the help during this study,
as well as to M. Rakhlin for the film deposition. This work
has been financially supported by. projects
Praxis/P/EEI/12183/1998 and POCTI/ESE/38689/2001.



REF
ERENCES




1
.

M. Vieira, M. Fernandes, J. Martins, P. Louro, A. Maçarico, R.
Schwarz, and M. Schubert, Mat. Res. Soc.
Symp. Proc., 609 (2000)
A14.2.

2
.

M. Vieira, M. Fernandes, J. Martins, P. Louro, A. Maçarico, R.
Schwarz, and M. Schubert, IEEE S
ensors Journal, Vol 1, No. 2 (2001)
158
-
167.

3
.

J. W. Horton, R. V. Mazza, and H. Dym, Proc. IEEE, 52 (1964) 1513
-
1519.

4
.

G. F. Amelio
,

Scientifc American, 230 (1974) 22
-
31.

5
.

M. Vieira, M. Fernandes, P. Louro, Y. Vygranenko, R. Schwarz, and,
M. Schubert
,
Mat. Res. Soc.
Symp. Proc.,

Vol. 664 (2001) A18.5.

6
.

R. Martins, A. Fantoni, and M. Vieira, J. of Non
-
Cryst. Solids
164&166

(1993) 671.

7
.

A Fantoni, M. Fernandes, P. Louro, R. Schw
arz and M. Vieira, In
Amorphous and Heterogeneous Silicon Thin Films
-

2001, Mat. Res.
Soc. Symp. Proc., (S. Francisco, USA), Vol. 664 (2001) A25.11.


8
.

C. Koch, M. Ito, M. Schubert, and J. H. Werner, Mat. Res. Soc.
Symp.
Proc. 575 (1999) 749.

9
.

P. Louro, M. Vieira, Yu.
Vygranenko, M. Fernandes, R. Schwarz, M.
Schubert, Ap
plied Surface Science, 184 (2001) 144
-
149.

10
.

M. Fernandes, Yu. Vygranenko, and M. Vieira, Applied surface
science, 184 (2001) 408
-
412.

V=
-
0.3 V

V=0 V

V=0.3 V