Post on 23-Oct-2014
KUMARAGURU COLLEGE OF
TECHNOLOGY
SIMULATION OF HUMAN NOSE
USING NEURAL NETWORKS
AUTHORS:
MANIVANNAN.P.S Manivannan.tn89@gmail.com 9894244556 GANESH RAJ.V Ganesh_stanite@yahoo.co.in 9943658100
ABSTRACT
ARTIFICIAL NEURAL NETWORK:
Artificial Neural Network (ANN) is an information processing technique that
Uses the biological nervous system , which includes the human brain to process some
data. It is composed of a large number of highly interconnected processing elements
(neurons) working in a very complex manner such that it is not noticed by either humans
or other computer techniques in unison, to solve specific problems. Neural networks have
the remarkable ability to obtain meaning from complicated data and can be used to
extract patterns and detect trends that are even unthinkable to the human mind.
ELECTRONIC NOSES:
In this paper we describe the various applications of neural networks and how
Neural networks are being implemented in designing Electronic/artificial noses. The
electronic nose is a small mechanical device that is capable of deciphering a wide
variety of smells, ranging anywhere from food to human breath to poisonous toxins.
An electronic nose is generally composed of a chemical sensing system (e.g., sensor
array or spectrometer) and a pattern recognition system (e.g., artificial neural network).
IMPLEMENTATION:
ANNs are systematically implemented in electronic noses. Electronic noses have several
applications in telemedicine. Telemedicine is the practice of medicine over long distances
via a communication link. The electronic nose would identify odours in the remote
surgical environment. These identified odours would then be electronically transmitted to
another site where an odour generation system would recreate them. Because the sense of
smell can be an important sense to the surgeon, telesmell would enhance telepresent
(surgery via communication links) surgery. Other applications of Electronic noses on
space & food industry are discussed. Who knows, maybe the future is just a sniff away!!
NEURAL NETWORKS: DEFINITION:
Neural Network (NN) is an
information processing paradigm that
is inspired by the way biological
nervous systems, such as the brain,
process information. It is a
sophisticated modeling technique
capable of modeling complex
functions. It is a nonlinear approach.
EASE OF USE:
Neural networks learn by example.
The user gathers up representative
data, invokes training algorithms to
auto learn the structure of data even
though user does not have heuristic
knowledge to select and prepare data
about neural networks, the level of
application of knowledge is much
lower than other nonlinear statistical
methods.
CHARACTERISTICS OF
NEURAL NETWORKS :
ROBUSTNESS AND FAULT
TOLERANCE:
The decay of nerve cells does not
seem to affect the performance
significantly.
FLEXIBILITY: The network
automatically adjust to a new
environment without using any
programmed instructions
ABILITY TO DEAL WTH
VARIETY OF DATA SITUATIONS:
The network can deal with
information that is fuzzy, noisy and
inconsistent.
COLLECTIVE COMPUTATION:
The network performs routinely many
operations in parallel and also a given
task in a distributed manner
HISTORICAL
BACKROUND: Neural network
simulations appear to be a recent
development. However, this field was
established before the advent of
computers, and has survived at least
one major setback and several eras.
Many important advances have been
boosted by the use of inexpensive
computer emulations. Following an
initial period of enthusiasm, the field
survived a period of frustration and
disrepute. During this period when
funding and professional support was
minimal, important advances were
made by relatively few researchers.
These pioneers were able to develop
convincing technology which
surpassed the limitations identified by
Minsky and Papert. Minsky and
Papert, published a book (in 1969) in
which they summed up a general
feeling of frustration (against neural
networks) among researchers, and was
thus accepted by most without further
analysis. Currently, the neural network
field enjoys a resurgence of interest
and a corresponding increase in
funding.
The first artificial neuron was
produced in 1943 by the
neurophysiologist Warren McCulloch
and the logician Walter Pits. But the
technology available at that time did
not allow them to do too much.
MERITS:
SPEED:
Biological Neural networks process
information in milliseconds range.
ANN in nanoseconds.
EFFICIENT: Able to detect more
chemicals than the number of sensors
used and allows less expensive
sensors.
DEMERITS:
FAULT TOLERANCE :
Information is distributed throughout
the network. Even if few connectors
are snapped information is still
preserved. But in computers if a chip
is corrupted information in the
memory cannot be retrieved
PROCESSING:
Neural networks can perform parallel
operations massively.
SIZE AND COMPLEXITY:
Number of neurons in brain is
estimated to be
around10^11.Computation is not
restricted to inside the soma whereas
in nn it is restricted
ARCHITECTURE:
NNs are composed of the
following elements:
Neuron (soma)
Inputs (dendrites)
Outputs of Neurons (axons)
Weights (synapses)
STRUCTURE OF BIOLIGICAL NEURON:
STRUCTURE OF ANN:
FUNCTIONING
All the weights of a NN comprise
its weight set, W = {w,v}.
The number of neurons per layer ,
number of hidden layers, and the
specified connections for each
layer comprise the network
architecture.
In the preceding figure, all of the
zeroth inputs to either the hidden
our output layer are referred to as
thresholds and are typically set to -
1.
The weights of a neural network
can be any positive or negative
value.
The input values are multiplied by
the weights that connect them to a
particular neuron.
Neurons take this weighted sum as
input and use an activation
function to compute the neurons
output.
The output of one neuron becomes
the input to another neuron
multiplied by a different subset of
weights.
The input coming into a neuron,
Hj, can be calculated as:
Hj = xi wij
Where xi represents the ith input and
wij represents the weight
connecting the ith input to the
jth hidden unit
The activation of Hj, f(Hj), can be
computed using a variety of
functions:
Here a sigmoid function is used
FORWARD PASS: Feed-forward
ANNs allow signals to travel only
one way from input to output.
There is no feedback i.e. the output of
any layer does not affect the same
layer
Hj = xi wij
Ok = f(Hj) vik (where
f is a sigmoid function)
Outk = f(Ok)
IMPLEMENTATION OF
ANN IN E-NOSE
ELECTRONIC NOSE:
Electronic noses are generally made
up of two main parts: a sensing system
and a pattern recognition system. In
the past,
gas chromatography and mass
spectrometry have been used as the
sensing system although these are
usually expensive and time
consuming. Today, the use of
chemical sensors has been established
to analyze odours. Essentially, each
odour leaves a characteristic pattern or
fingerprint of compounds. Known
odors can be used to build up a
database to train a pattern recognition
system. One possibility is to have a
sensor for every chemical, though this
would be costly since there are so
many different chemicals. The answer
is in artificial neural networks
(ANNs). ANN are able to detect more
chemicals than number of sensors it is
H1
H2
H3
O1
H0 x0
x1
x2
x3
w01
w12 w13
w21
w22
w23
w31 w32
w33
w02 w03
w11
v01
v11
v21
v31
Neuron (Soma) Input (dendrite) Output (Axon) Weight (Synapse)
Out1
utilizing. ANNs also allow for less
selective and therefore less expensive
chemical sensors. The artificial neural
networks are trained to distinguish
certain odours from certain chemical
combinations. Pattern recognition is
gained through giving the network
known odours and classifying them
with a signature. Then the nose is
tested to see how well the ANN has
learned. The results can be adjusted
through experimentation. The sensors
basically measure the change in
voltage due to the presence of certain
chemicals. The chemicals in the air
change the oxygen content over the
sensors, which are electronic circuits.
By changing the oxygen content, the
resistance across the sensor is changed
which can be measured as a voltage
drop from the normal or standardized
conditions. This analog signal must
then be translated into a digital signal
by an A/D converter in order for the
computer to understand the
information. The number of odour
signatures the system can recognize
depends on the number of sensors
used and the number of grey levels in
the convertor. The maximum signature
number is given by gn, where n is the
number of sensors and g is the number
of grey levels. A 10-bit converter has
a grey level value of 1024, so an array
of three sensors could yield over a
billion different signatures.
Unfortunately, the actual number is far
below this value.
SENSING SYSTEM:
There are two types of sensors namely
>polymer based sensors
>metal oxide sensors.
POLYMER BASED SENSORS: The inside of a polymer-based
electronic nose is composed of various
types of polymer films. These
polymer films are made up of an
insulator and a filler. The polymer
insulator actually absorbs some of the
gas molecules. The substance that is
found inside of the insulator is a
conducting polymer with a certain
resistance. Each film in the electronic
nose is made of a different insulator
and a different conducting polymer.
The resistance of the polymer films
before exposure to a gas is initially
recorded. Then, as a gas passes
through the electronic nose, it is
actually absorbed into the polymer
films. Upon absorption of the gas, the
polymer insulators swell. This affects
the conducting polymer contained
inside of the insulator by limiting the
number of connected pathways
throughout the conductor . The larger
the insulator swells, the fewer the
number of connections in the polymer
conductor, which decreases the
resistance of each polymer film to a
degree.
Since each film in the electronic nose
is made of a different insulator and
conducting polymer, the final
resistance of each film is different.
The combination of the ultimate
resistances forms the fingerprint
needed for the identification of the
gas. The films are connected to an
artificial neural network, which
contains pre-programmed fingerprints
of odors. If the smell is not
programmed into the machine, it will
not be able to identify it. Using the
programmed fingerprints, the
polymer-based electronic nose is then
able to identify the gas. This data is
finally sent to a computer, which
displays the data determined by the
electronic nose.
METAL OXIDE SENSORS:
Metal oxide-based electronic noses
also contain several main parts, but
they are mostly different than those of
the polymer-based electronic nose. It
must also contain a power supply in
order to run the machine. Instead of
an air pump, the metal oxide-based
electronic nose contains a sampling
chamber where the sample to be
analyzed is placed. A pump and fan
system then wafts air from the sample
through the electronic nose and over
its sensors.
These sensors are contained in
chambers located inside of the
electronic nose. In order to work
properly, the chambers must be kept at
an elevated temperature, so a heater is
also attached to the electronic nose.
Each chamber, constructed of a non-
reactive material, which does not
affect the metal oxide sensors,
contains a various number of metal
oxide sensors, as well as one
integrated sensor. This integrated
sensor is sensitive to changes in
humidity and temperature and records
the changes relative to the sample .
Therefore, if a drastic change occurs,
the experimenter will be able to
interpret the data accordingly.
The sensors contained in a metal
oxide-based electronic nose vary in
number and composition according to
the electronic nose’s purpose.
Determining the number and
composition of sensors needed comes
down to understanding exactly how
the sensors operate. Each sensor is
composed of a ceramic substance
coated with a different semi-
conductive metal oxide film. When
exposed to a gas, the surfaces of the
films chemically react with the gas,
and an electron transport takes place.
This translates into a change in the
sensors’ resistances. Since each
surface is covered in a different metal
oxide, each reacts differently to the
gas and therefore, has a different
change in resistance.
. An electronic nose designed to
detect a simple gas will only have a
few sensors, while one designed to
detect a more complex gas will require
many more sensors.
Once the sensors react to the
gas, a fingerprint of the odor is then
made up of the combination of the
sensors’ changes in resistance. This
fingerprint is sent to a data acquisition
device, which is hooked up to the
electronic nose. The device then
translates the fingerprint into the
actual identity of the gas and sends
this information to a computer.
WHICH IS BETTER?
It is hard to definitely say which type
of electronic nose is better because it
depends on how it is being applied
In general, though, the polymer-based
electronic nose seems to be superior to
the metal oxide-based electronic nose.
The metal oxide-based
electronic nose surpasses the polymer-
based in only a few limited areas. The
metal oxide-based electronic nose
shows a high repeatability due to the
fact that its sensors do not change
shape to record the electrical data.
They also show a very high sensitivity
to substances, especially organic
compounds.
Beyond these few areas, the
metal oxide-based electronic nose falls
short. Since it has to be kept at a high
temperature, it can be quite a hassle to
use. Its sensors also demonstrate a
very low selectivity, and because of
this, they must be used in arrays. The
sensors are not sensitive enough to be
used on their own.
Polymer-based electronic
noses greatly surpass the metal oxide-
based electronic noses. They are more
convenient because they can be used
at any temperature. They also
demonstrate lower power consumption
as well as a faster response time.
The polymer sensors are what
really make this electronic nose better.
The fact that they are made of
polymers offers unlimited possibilities
for the types of sensors as well as the
arrangement of arrays. Not only do
they make diverse arrays, the polymer
sensors have such a high selectivity
that they can be used individually as
well. Unlike the metal oxide sensors,
polymer sensors also show no
sensitivity to water, increasing the
environments in which this type of
electronic nose can be used. The
sensors are also very easy to make and
can be done so at a lower cost than the
metal oxide sensors. Finally, they
show a durability of at least six
months, which is long for this type of
new technology . Research
overwhelmingly suggests that
polymer-based electronic noses
surpass metal oxide-based electronic
noses in almost every category. While
some industries do prefer the metal-
oxide based electronic nose, most
choose the polymer-based for its
convenience, reliability, and accuracy.
PRE-TRAINING PROCESS:
Each chemical vapor presented to
the sensor array produces a
signature or pattern characteristic
of the vapor.
By presenting many different
chemicals to the sensor array, a
database of signatures is built up.
This database of labeled signatures
is used to train the pattern
recognition system.
The goal of this training process is
to configure the recognition
system to produce unique
classifications of each chemical so
that an automated identification
can be implemented
An Example to illustrate the above
training procedure:
Assume that we want a network to
recognise hand-written digits. We
might use an array of, say, 256
sensors, each recording the presence
or absence of ink in a small area of a
single digit. The network would
therefore need 256 input units (one for
each sensor), 10 output units (one for
each kind of digit) and a number of
hidden units. For each kind of digit
recorded by the sensors, the network
should produce high activity in the
appropriate output unit and low
activity in the other output units. To
train the network, we present an image
of a digit and compare the actual
activity of the 10 output units with the
desired
activity. We then calculate the error,
which is defined as the square of the
difference between the actual and the
desired activities. Next we change the
weight of each connection so as to
reduce the error. We repeat this
training process for many different
images of each different images of
each kind of digit until the network
classifies every image correctly.
PATTERN RECOGNITION
SYSTEM:
An important application of neural
networks is pattern recognition.
Pattern recognition can be
implemented by using a feed-forward
(figure 1) neural network that has been
trained accordingly. During training,
the network is trained to associate
outputs with input patterns. When the
network is used, it identifies the input
pattern and tries to output the
associated output
pattern. The power of neural networks
comes to life when a pattern that has
no output associated with it, is given
as an input. In this case, the network
gives the output that corresponds to a
taught input pattern that is least
different from the given pattern.
CONTROL ALGORITHM:
The Electronic nose consists of two
objects to be controlled namely the
Control of inhalation pump
Control of valve responsible
for phase measurement.
The control program of an Electronic
nose is implemented using these steps:
Turn on the pump
collect one sample from each
active sensor
send the samples to the
display
If all samples are collected
goto step5 else goto step 2.
Turn off the pump
Data acquisition from the nose
is finished.
The control program is basically
structured as an endless loop within a
main function . A switch statement tat
checks the variable “com” ,which is
used in the loop to direct the program
to the correct action decided by the
input character.
The variable “com” gets its character
from a receive – buffer with the help
of
an interrupt service routine. The basic
structure of program is as follows:
Void main(void)
{
While(1)
{
com = getchar(); // wait
for serial interface
switch(com)
{
Case’!’:
break;
case’L’:
default:
break;
} }}
MAJOR APPLICATIONS:
ELECTRONIC NOSE IN SPACE:
In the space station, astronauts are
surrounded by ammonia. It flows
through pipes, carrying heat generated
inside the station outside to space.
Ammonia helps keep the station
habitable.
But it's also a poison. And if it leaks,
the astronauts will need to know
quickly. Ammonia becomes
dangerous at a concentration of a few
parts per million (ppm).
Hense electronic noses comes into
application in space station. Here the
Enose consists of 16 different polymer
films.These films are capable of
conducting electricity and are used to
detect the gases.
ELECTRONIC NOSES IN TELEMEDICINE:
Because the sense of smell is an
important sense to the physician, an
electronic nose has applicability as a
diagnostic tool.
An electronic nose can examine
odours from the body (e.g., breath ,
wounds, body fluids, etc.) and identify
possible problems. Odours in the
breath can be indicative of
gastrointestinal problems, sinus
problems, infections, diabetes, and
liver problems.
Infected wounds and tissues emit
distinctive odors that can be detected
by an electronic nose. Odours coming
from body fluids can indicate liver and
bladder problems.
ELECTRONIC NOSES IN FOOD INDUSTRY:
Applications of electronic noses
in the food industry include
Quality assessment in food
production .
inspection of food quality by
odour.
control of food cooking processes.
inspection of fish.
monitoring the fermentation
process.
verifying if orange juice is natural.
Monitoring food and beverage
odours.
inspection of beverage containers.
E- NOSE IN SEWAGE:
Each stage of a sewage treatment
process emits odor causing
compounds and these compounds may
vary from one location in sewage
treatment works to another. In order to
determine the boundaries of legal
standards reliable and efficient odor
measurement methods need to be
measured. An E-NOSE equipped with
12 different poly pyrrole sensors is
used for characterizing sewage odour.
CONCLUSION:
The computing world has a lot to
gain from neural networks. Their
ability to learn by example makes
them very flexible and powerful.
Perhaps the most exciting aspect of
neural networks is the possibility that
some day 'concious' networks might
be produced. There is a number of
scientists arguing that conciousness is
a 'mechanical' property and that
'concious' neural networks are a
realistic possibility.
Judging by its popularity in
current trade and scientific magazines
and its various practical uses, the
future of ANN in electronic nose
looks very bright. It is being
introduced into the medical world as a
device that can actually identify a
disease from a sample of the patient’s
breath. So far, the electronic nose has
a wide variety of uses, and its
possibilities are literally endless. In
the near future, the electronic nose
could very well become a part of
everyday life.
Finally, we would like to state that
even though neural networks have a
huge potential we will only get the
best of them when they are integrated
with computing, AI, fuzzy logic and
related subjects.
BIBLIOGRAPHY:
“Articifial Neural
networks”,B.YEGNANARAYA
NA.
B.S. Hoffheins, Using Sensor
Arrays and Pattern Recognition
to Identify Organic Compounds.
MS-Thesis,
B.S. Hoffheins, Using Sensor
Arrays and Pattern Recognition
to Identify Organic Compounds.
MS-Thesis,