real time monitoring odor sensing system using omx

TRANSMISSION OF THE OLFACTORY INFORMATION BEKIR KARLIK AND YOUSIF AL-BASTAKI The College of Information Technology University of Bahrain P.O. Box: 32038, Sukhair Campus KINGDOM OF BAHRAIN [email protected] and [email protected]

Abstract: - There have been many works for odor recognition using different sensor arrays and pattern recognition techniques in last decades. Although an odor is usually recorded utilizing language expression, it is too difficult for laymen to associate actual odor with that expression. The odor sensing system should be extended to new areas since a neural network recognizes its standard style where the output pattern from multiple sensors with partially overlapped specificity. In this study two works were realized. The first we have developed odor sensing system with the capability of the discrimination among closely similar 10 different perfumes odor patterns and proposed a real time classification method using a handheld odor meter (OMX-GR sensor) and neural network. The other, we have developed telemedicine odor-sensing system with the capability of the discrimination among different bad breath odor patterns obtained sugar diabetic persons.

1. Introduction This following of Holy Koran has been giving a clue to us about transmission of smelling information

"‫تفندون‬

‫"و لما فصلت العير قال أبوهم إني ألجد ريح يوسف لوال أن‬

)94 ،‫(يوسف‬ "When the caravan left (Egypt) their father said: I do indeed scent the presence of Yusuf: Nay, think me not a dotard" [1].

The goal of much of the research regarding the olfactory system is to understand how individual odors are identified. Many researchers have produced mathematical models of the olfactory system [2-6]. These models often include simulations of the neurobiological information processing systems.

Artificial Neural Networks (ANN) is an abstract simulation of a real nervous system that contains a collection of neuron units communicating with each other via axon connections. Such a model bears a strong resemblance to axons and dendrites in a nervous system. Artificial Neural Networks are programs designed to simulate the way a simple biological nervous system is believed to operate. A neural network is an information processing system that is non-algorithmic, non-digital, and intensely parallel. In which our senses provide us with rapid information about our environment. And these senses respond to a variety of stimuli. The sense of touch is stimulated by the pressure of physical contact with an object [2]. Hearing responds to rapid fluctuations in air pressure. Sight is produced by electromagnetic radiation falling on the retinas of our eyes. Two of our senses respond to the chemical nature of our surroundings: taste and smell. Because they depend on chemical interactions, these two senses are called chemoreception. Taste is called contact chemoreception, because to experience the flavour of something, we must come into contact with it. Smell is remote chemoreception, for we can sense the odour of an object at a distance.

ANN, which has been used to analyse complex data and for pattern recognition, are showing promising results in chemical vapour recognition [7-8]. When an ANN is combined with a sensor array, the number of detectable odors is generally greater than the number of sensors. While the inclusion of visual, aural, and tactile senses into virtual reality systems is widespread, the sense of smell has been largely ignored [9-10]. We have studied a chemical vapour sensing system for the automated identification of chemical vapours (smells). Our prototype chemical vapour sensing system is composed of an array of chemical sensors (usually gas sensors) coupled to an artificial neural network. The artificial neural network is used in the recognition and classification of different smells and is constructed as a standard multilayer feed-forward network trained with the back propagation algorithm. When a chemical sensor array is combined with an automated pattern identifier, it is often referred to as an electronic or artificial nose [11]. This report was designed with the following structure:

Chapter 1: Introduction; Brief description of the idea of the project, its problem and the objectives behind it.

Chapter 2: Biological and Artificial Nose. Describing the idea of the Artificial (or electronic) Nose, it’s content and the major part of it, which is the sensor and how does it work.

Chapter 3: Artificial Neural Network (ANN); Since this paper is about The Electronic Nose which is built on the idea of ANN, it was recommended to devote a whole chapter explaining the idea behind this new technology.

Chapter 4: The Transmission System and Applications; It explains the core idea of the Tele-smell and telemedicine, which is the main objective of this paper. It contains practical applications used to enable the transmission of the olfactory information and artificial nose virtually.

2. The Biological and Artificial Nose 2. 1. The Olfactory System The olfactory information is processed in both the olfactory bulb and in the olfactory cortex. Figure 1 illustrates the main information processing structures within the brain. The olfactory bulb performs the signal preprocessing of olfactory information including recoding, re-mapping, and signal compression. The olfactory bulb also handles cases where an odor presented for a long time produces habituation. The olfactory cortex performs pattern classification and recognition of the sensed odors. Once identified, odor information is transmitted to the hippocampus, limbic system, and the cerebral cortex. The connection to the hippocampus explains why odor can sub-consciously evoke memories. Conscious perception of the odor and how to act on the odor takes place in the cerebral cortex. It is interesting to consider that the mammalian epithelium contains from approximately 1 million sensory neurons in the mouse, to 10 million sensory neurons in the human, to 100 million sensory neurons in the pig [2].

Fig. 1 The major processes of the olfactory system

In the olfactory system, hundreds of different odorant receptors are used in a combinatorial fashion to encode the identities of thousands of odorous chemicals. Studies using receptor genes as molecular and genetic tools have revealed how these combinatorial codes are represented in the nose, olfactory bulb, and olfactory cortex to ultimately generate diverse odour perceptions [2-7].

The Olfactory nerves react as other nerves in the body do, responding to electrical signals and impulses and dispatching information to the rest of the body. The sense of smell (olfaction) is both a very simple and a very complex sense. It is simple because relatively few cells are involved in detecting odours. In humans, the olfactory sensors are located at the top of the nasal passages, just below and between the eyes. Each passage contains a small area (about 2.5 cm2) containing roughly 50 × 106 receptor cells. Each of these cells is connected through a single synapse (junction between nerve cells) directly to the brain [2]. Of all our senses, the sense of smell is the most intimately connected with the brain. In spite of this, the sense of smell is very complex in how it functions.

The mechanism by which the odour receptor cells interact with odour-causing molecules is still unknown, but studies of odours and the structure of the odour-causing molecules have revealed some correlations.

2.2. The Biological Nose The mammalian olfactory system uses a variety of chemical sensors, known as olfactory receptors, combined with signal processing in the olfactory bulb and automated pattern recognition in the olfactory cortex of the brain. No one-receptor type alone identifies a specific odour. It is the collective set of receptors combined with pattern recognition that results in the detection and identification of each odour. Figure 2-(a) illustrates the major operations of the mammalian olfactory system [7]. The operations can be broken into sniffing, reception, detection, recognition, and cleansing.

Figure 2 a) illustrate the major components of the olfactory system b) Illustrate the major components of the electronic nose

It can be seen Fig.2-a, through sniffing, odor molecules arrive at the olfactory receptors stimulate an electro-chemical response that is transmitted to the olfactory bulb and ultimately the olfactory cortex for identification. The olfaction process begins with sniffing, which brings odorant molecules from the outside world into the nose. With the aid of turbinate (bony structures in the nose which produce turbulence), sniffing also mixes the odorant molecules into a uniform concentration and delivers these

molecules to the mucus layer lining the olfactory epithelium in the upper portion of the nasal cavity. Next, the odorant molecules dissolve in this thin mucus layer which then transports them to the cilia (hair like fibers) of the olfactory receptor neurons. The mucus layer also functions as a filter to remove larger particles [2].

Reception involves binding the odorant molecules to the olfactory receptors. These olfactory receptors respond chemically with the odorant molecules. This process involves temporarily binding the odorant molecules to proteins that transport the molecules across the receptor membrane. Once across the boundary, the odorant molecules chemically stimulate the receptors. Receptors with different binding proteins are arranged randomly throughout the olfactory epithelium. The chemical reaction in the receptors produces an electrical stimulus. These electrical signals from the receptor neurons are then transported by the olfactory axons through the cribiform plate (a perforated bone that separates the cranial cavity from the nasal cavity within the skull) to the olfactory bulb (a structure in the brain located just above the nasal cavity). From the olfactory bulb, the receptor response information is transmitted to the olfactory cortex where odor recognition takes place. After this, the information is transmitted to the limbic system and cerebral cortex [5]. There are no individual olfactory receptors or portions of the brain that recognize specific odors. It is the brain that associates the collection of olfactory signals with the odor. Finally, in order for the nose to respond to new odors, the olfactory receptors must be cleansed. This involves breathing fresh air and the removal of odorant molecules from the olfactory receptors.

2.3. The Artificial Nose Artificial noses are much simpler than almost all biological olfactory systems and detect only a small range of odours. However, for many potential Tele-smell applications in the near future, a predetermined and limited set of odours is likely. Thus, it is likely an electronic nose will be a key component in an olfactory input to a Tele-present virtual reality system [7].

Artificial noses (or Electronic noses) are being developed as systems for the automated detection and classification of odours, vapours, and gases. 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) (see Figure 2-b). An artificial nose is composed of a chemical sensing device and an automated pattern recognition system. This combination of broadly tuned sensors coupled with sophisticated information processing makes the electronic nose a powerful instrument for odor analysis. The sensing system can be an array of chemical sensors where each sensor measures a different property of the sensed chemical, or it can be a single sensing device (e.g., gas chromatograph, spectrometer) that produces an array of measurements for each chemical, or it can be a hybrid of both. Each odorant or volatile compound presented to the sensor array produces a signature or characteristic pattern of the odorant.

An artificial neural network (ANN) is an information-processing paradigm that was inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurons) working in unison to solve specific problems. ANN, like people, learns by example. An ANN is configured for an application such identifying chemical vapors through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurons. This is true of ANN as well. For the artificial nose, the ANN learns to identify the various chemicals or odors by example.

Although each sensor is designed for a specific chemical, each responds to a wide variety of chemical vapours. Collectively, these sensors respond with unique signatures (patterns) to different chemicals. During the training process, various chemicals with known mixtures are presented to the system. ANN, which have been used to analyse complex data and for pattern recognition, are showing promising results in chemical vapour recognition. When an ANN is combined with a sensor array, the number of detectable odours is generally greater than the number of sensors. Less selective sensors are generally less expensive that can be used with this approach. Artificial noses that incorporate ANN have been demonstrated in the following applications [8-22]: 

Quality control in the food industry



Quality control of packaging material



Medical diagnostics



Environmental monitoring



Perfume and aroma industry



Control of beverages, e.g. wine and beer



Tobacco industry



Coffee industry



Assessment of car interiors

In this study three different practical examples about electronic nose were presented to use in the different areas (industrial and medical).

3. Artificial Neural Networks (ANN) 3.1. Introduction to ANN An Artificial neural network (ANN) is a paradigm of information processing that was inspired by the way biological nervous systems, such as the brain, process information. The key element of this paradigm is the novel structure of the information processing system. It is not a computer in the sense we think of them today, nor is it programmed like a computer. Instead, it consists of a number of very simple and highly interconnected processors called neurons, which are the analogue of the biological neural cells, or neurons, in the brain. The first fundamental ANN modelling proposed in 1943 by McCulloch and Pitts in terms of a computational model of "nervous activity". The McCulloch-Pitts neuron is a binary device and each neuron has fixed threshold logic. This model leads the works of John von Neumann, Marvin Minsky, Frank Rosenblatt, and many others [23].

The basic unit of an artificial neural network is the neuron. Each neuron receives a number of inputs, multiplies the inputs by individual weights, sums the weighted inputs, and passes the sum through a transfer function, which can be, e.g., linear or sigmoid (linear for values close to zero, flattening out for large positive or negative values). An ANN is an interconnected network of neurons. The input layer has one neuron for each of the sensor signals, while the output layer has one neuron for each of the different sample properties that should be predicted. Usually, one hidden layer with a variable number of neurons is placed between the input and output layer [24]. During the ANN training phase, the weights and transfer function parameters in the ANN are adjusted such that the calculated output values for a set of input values are as close as possible to the known true values of the sample properties. It is composed of a large number of highly interconnected processing elements (neurons)

working in unison to solve specific problems for this study. It consists of three interconnected layers of neurons (Fig. 3).

Fig. 3 Schematic of an artificial neural network

The parameters of the neurons are chosen through a minimization of the output error for a known training set. ANN, like people, learns by example. An ANN is configured for an application such identifying chemical vapours through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurons. This is true of ANN as well. For the artificial nose, the ANN learns to identify the various chemicals or odours by example. Another advantage of the parallel processing nature of the ANN is the speed performance. During development, ANN is configured in a training mode. This involves a repetitive process of presenting data from known diagnoses to the training algorithm. This training mode often takes many hours. The payback occurs in the field where the actual odour identification is accomplished by propagating the data through the system, which takes only a fraction of a second. Since the identification time is

similar to the response times of many sensor arrays, this approach permits real-time odour identification.

Recently, it has been shown that neural networks have abilities to solve not only artificial noses also various complex problems 23-24. Several ANN configurations have been used in electronic noses including back propagation-trained, feed-forward networks; Kohonen’s self-organizing maps (SOMs); Learning Vector Quantizers (LVQs); Hamming networks; Boltzmann machines; and Hopfield networks [8-22]. In this study a Multi-Layered neural network with back-propagation training algorithm, which has generalized delta rule learning is used.

3.2 The Architecture of Multi Layer Perceptron and Back-Propagation Algorithm On the other hand, the multi-layered feed-forward networks has a better ability to learn the correspondence between input patterns and teaching values from many sample data by the error backpropagation algorithm 24. Therefore, in this paper we used a three-layered feed forward neural network and trained it by error back-propagation. The software of ANN is written and employed back-propagation in a supervised learning paradigm in which the generalized delta rule was used in adjusting the weight values. Fig. 3 shows a general structure of ANN. Each layer is fully connected to the previous layer, and has no other connection. The output Oj of each unit j is defined by, Oj = (netj), netj =  w ji Oi   j i

( i preceding layer)

(2)

where Oi is the output of unit i, wji is the weight of the connection from unit i to unit j, j is the bias of unit j, i is a summation over every unit i whose output flows into unit j, and (x) is a monotonously increasing function. The computing neurons (hidden and output layers) have a non-linear transfer function. In practice, a logistic activation function (sigmoid function) (x) = 1/(1+exp(-x)) is used.

When the set of m-dimensional input patterns ip = (ip1 , ip2 , .... , ipm ) ; pP where P denotes set of presented patterns, and their corresponding desired n-dimensional output patterns tp = (tp1 , tp2 , .... ,

tpm ) ; pP are provided, the neural network is trained to output ideal patterns as follows. The squared error function Ep for a pattern p is defined by Ep 

 1 2   (t pj  o pj )  2  j output 

(3)

tpj : target (desired) value, opj : actual network output value.

The purpose is to make E  p Ep small enough by choosing appropriate wji and j. To realize this purpose, a pattern p  P is chosen successively and randomly, and then wji and j are changed by p wji = -  (Ep /  wji )

(4)

p j = -  (Ep / j )

(5)

where  is a small positive constant. By calculating the right hand side of (4) and (5), it follows that p wji =  p j Op i

(6)

p j =  p j

(7)

where

p j

 (net j )(t pj  O pj ) (when j belongs to the output layer. )  =    (net )  w  j k kj pk (otherwise) 

(8)

Note that k in the above summation represents every unit k in the layer following the layer of j (unit j). In order to accelerate the computation, the momentum terms are added in (6-7), p wji (n+1) =  p j Opi +  p wji (n)

(9)

p j (n+1) =  p j +  p j (n)

(10)

where n represents the number of learning cycles, and  is a small positive value.

4. The Transmission System and Applications

4.1. Tele-Smell System The artificial nose described in this works is adequate for demonstrating the concept of Tele-smell and olfactory input in a virtual reality environment. The ultimate goal of our research will be to demonstrate Tele-smell with odours of importance in Tele-present surgery and Tele-present battlefield surgery. This will involve the construction of an automated odour sensing system (electronic nose) that can identify odours generated by the human body (e.g., bile, urine, blood, etc.), integration of the electronic nose with the odour generation system, and demonstration of Tele-smell in Tele-surgery. Figure 4 illustrates a possible system for demonstrating Tele-smell. It is composed of an odor identification system (e.g., electronic nose), transmission channel, and an odor regeneration system. To date, we have only demonstrated the odor identification part. However, we are currently preparing for a demonstration of Tele-smell. The goal of this demonstration is to show the concept of Tele-smell and olfactory input in a virtual reality environment.

Fig. 4 Real-time monitoring odour sensing and Tele-smell system

While the inclusion of visual, aural, and tactile senses into virtual reality systems is widespread, the sense of smell has been largely ignored. We have studied a chemical vapour sensing system for the automated identification of chemical vapours. Our prototype chemical vapour sensing system is

composed of an array of chemical sensors (usually gas sensors) coupled to an artificial neural network. The artificial neural network is used in the recognition and classification of different odours and is constructed as a standard multi layer feed-forward network trained with the back-propagation algorithm. When a chemical sensor array is combined with an automated pattern identifier, it is often referred to as an electronic or artificial nose.

Odours molecules arrive at the chemical sensor array stimulate an electrical response that is transmitted to the pattern recognition system and ultimately to an output display or actuation. Odour (or olfactory) information can also transmit using computer network system.

The prototyped ANN was constructed as a multi layer feed-forward network and was trained with the back-propagation of error algorithm by using a training set from the sensor database. This prototype was initially trained to identify odours of 10 different perfumes. This system allows users to obtain the desired data from a particular odorant (perfumes). There are two ways to obtain data by using a handheld odour meter (OMX-GR sensor):  Real Time Sampling Data  Memory Sampling Data The system mainly contains three forms: 1. The first from, shown in Fig. 5, allows user to choose among two buttons in which when the user clicks on any one of the buttons an open dialog box well appears (shown in Fig. 6), asking the user to enter the name of the file.

Fig. 5 The main form in the sensor program

Fig. 6 Open dialog box

Fig. 7 The Real time sampling form

2. Real Time sampling form (the second form) shown in Fig. 7. It appears when the user chooses the real time sampling data button from the first from. 3. Memory sampling data form (the third form) shown in Fig. 8. It appears when the user chooses the Memory sampling data button from the first from.

Fig. 8 The Memory sampling data form

Finally this is the Artificial Neural Network System, which classifies the data and tests them. The system asks the user to enter some values and input file name, after learning session the system well create four new file, assume that the input file name is first_learning.dat , then it well create the following files: 

first_learning_w.dat: this file contains the weights.



first_learning_v.dat: this file contains the value.



first_learningy.cns: this file contain the



first_learning.err: this file contains the error.

And for the output generation it well creates: 

first_learninght.dat: this file contains the output of the testing session.

At the beginning the program well ask the user to enter L for learning, O for output generation or 1 to continue from old weights file. 1- If the user chooses learning, the program well asks you to enter the task name that contains data. a. Then the user should enter the number of features in each input pattern, which in our case are 26x10 (each odour contain 26 samples). b. Then the user should enter the number of output units, which in our case 10 outputs (10 odour samples). For medical application 2 outputs such as normal and abnormal. c. Continued by entering the number of input samples, which are also 10 in our case. d. The program well search for the file that the user entered & if it found it then it well ask the user if he\she wants to take a look of the data in the file, just to read by entering yes or no. e. Then the user should enter the momentum rate value, and it’s by default 0.9 and followed by the learning rate Alfa, which is by default 0.7. f. Enter the maximum number or iteration, by default its 1000, but its butter to enter a number that is grater than 1000. g. Then the program well asks the user to put the number of hidden layers, and a number of layer units for each layer. h. The last thing before starting the learning session, the program well ask the user if she\he wants to create an error file or not, if yes press 1 if no press 0.

The learning phase well start and the program well ask the user to wait until it finishes the training (see Fig. 9). 2- If the user chooses the testing, the program well asks the user if she\he wants to work on a different learning task or not. After that it well ask the user for the testing input file name, if the user enters a correct file name it well asks the user to enter the number of patterns for processing.

Fig. 9 The ANN program in the learning session

4.2. Bad Breathe Diagnosis System for Telemedicine Recently the use of smell in clinical diagnosis has been rediscovered due to major advances in odour sensing technology and ANN [19-22]. It was well known in the past that a number of infectious or metabolic diseases could liberate specific odours characteristic of the disease stage. Later chromatographic techniques identified an enormous number of volatiles in human clinical specimens

that might serve as potential disease markers. ANN has been employed in several areas of medical diagnosis, including rapid detection of tuberculosis, Helicobacter pylori, Infection or cancer of nose or sinuses or teeth, urinary tract infections, sugar diabetic, gastric, pulmonary and urine diagnosis etc. Preliminary results have demonstrated the possibility of identifying and characterizing microbial pathogens in clinical specimens. Initial clinical tests have shown that it may be possible in the near future to use electronic nose technology not only for the rapid detection of diseases such as peptic ulceration, urinary tract infections, and sugar diabetic but also for the continuous dynamic monitoring of disease stages.

In this project, we have developed telemedicine odour-sensing system with the capability of the discrimination among different bad breath odour patterns obtained sugar diabetic persons. This proposed a real time classification method has two main parts, which are a handheld odour meter (OMX-GR sensor) for obtaining data and ANN for classification (or diagnosis). A standard multi layered perceptron (MLP) feed-forward network trained with the back-propagation algorithm was used in the diagnosis belong to collecting data of bad breaths of sugar diabetic patients and normal breaths of different adults. This primary study has two main stages. The first stage is real time diagnosis of bad breath; the second is transmission of this information using telemedicine system. Fig. 10 illustrates a prototype based on electronic nose diagnosis system, which is used to identify bad breath from patient who has sugar diabetic illness problem.

Fig. 10 Real-time diagnosis of bad breaths and Transmission

During operation, the sensor array “smells” a breath odour, the sensor signals are digitised and fed into a computer, and the ANN (implemented in software) then identifies the chemical. This identification time is limited only by the response of the chemical sensors, but the complete process can be completed within seconds. The proposed ANN program is very useful for real-time odour recording and odour recognition system. The second stage of this study is to transmitting recognition data to the other areas (such as clinic or hospital) to be checked the results by a medical doctor.

Fig. 11 Transmission of the identified data

This system is built using VB codes that present the main interface of the system. The system enables users to connect remotely to another computer and transfer data virtually. Users are able to interact with the system through several buttons and so they are able to open a data file, obtain data, save the file, prepare the file to be classified, prepare the file for testing, connect to the remote point and disconnect from remote point, as shown in the Fig. 11. Buttons: -

Obtaining Data: this button links the user to the sensor system, which enables obtaining new data.

-

Classification: this button links the user to the ANN system, which enables classifying the chosen data.

-

Testing: this button links the user to the ANN system, which enables testing the chosen data.

-

Open Data File: this button opens the open dialog box, which enables the user to select a file.

-

Save Data File: this button enables the user to save data.

-

Prepare Data for Classification: this button is to prepare the file to be ready for classification by changing the (,) to (.) in the file and normalize the data by dividing the numbers by 100. Then add the desired value matrix according to the number of reading.

-

Prepare Data for Testing: this button is to prepare the file to be ready for classification by changing the (,) to (.) in the file and normalize the data by dividing the numbers by 100.

-

Connect to Remote Point: this button enables users to connect to another PC, which are connected together with a point-to-point connection.

-

Disconnect From Remote Point: this bottom used to disconnect the connection with the other Pc.

-

Clear Log: this button is to clear the action log that appears in the form.

-

Save Log: this form is to save the action log that appears in the form.

The proposed ANN program is very useful for real-time biomedical odour record and bad breathe recognition system, which has a various types of breath samples. The software program was developed by C++ medium level programming language

4. 3. Conclusions and Results This proposed works presents only to diagnosis sugar diabetic illness. The recognition rate was above 90 %. Depending on used ANN architecture, optimum learning rate (), and momentum coefficient () were found as 0.7 and 0.9 respectively. Training the ANN System using the data we have collected during our study of the Electronic Nose resulted in the following out put of error. It can be seen in Fig. 12, the number of hidden layers was fixed to one hidden layer, and the number of nodes (or units) in that hidden layer was changed several times. Also the iteration number was fixed to 5000 iterations. These results (the output error) were drawn together with the number of nodes in the hidden layer in a curve (see Fig.12).

Fig. 12 Error according to number of nodes for one hidden layer

If some patients have different bad-breaths such as infections from their teeth, nose, tonsil, stomach, oesophagus or Infection of cancer of nose or sinuses etc., which sometimes effects to recognize sugar diabetic problem. In future we are planning to improve this study.

The other problem for an odour recorder, which records an odour and reproduces it any time, has been studied. So, the odour recorder for recording the dynamical change of odour was studied since the odour in atmosphere is always changing. Due to the limitations of current technology, many ANN based artificial noses have less than 20 sensing elements and less than 100 neurons.

These systems are designed for specific applications with a limited range of odours. Systems that mimic more of the functionality of the human olfactory system will require a much larger set of sensing elements and a larger ANN. During operation, the sensor array “smells” an odour, the sensor signals are digitised and fed into a computer, and the ANN (implemented in software) then identifies the chemical. This identification time is limited only by the response of the chemical sensors, but the complete process can be completed within seconds.

Recently, some scientists have been working on the transmission of smelling system. Unfortunately they couldn’t manage to develop an opposite sensor, which converts the electrical signals to the odour signals. In the future, if it will be developed, we can smell our son’s odour like prophet Jacob (AS).

References [1] Holy Koran, section of Koran (Sure): Yusuf, verse of the Koran: 94 [2] Z. Li and J.J. Hopfield, Modelling the Olfactory Bulb and its Neural Oscillatory Processing, Biological Cybernetics, 61, 1989, pp. 379-392. [3] Y. Yao and W.J. Freeman, Pattern Recognition in Olfactory Systems: Modeling and Simulation. In: Proceeding of the 1989 International Joint Conference on Neural Networks (IJCNN'89). IEEE Press, Piscataway, NJ, 1989, Vol. 1, pp. 699-704. [4] B.S. Hoffheins, Using Sensor Arrays and Pattern Recognition to Identify Organic Compounds. MS Thesis, The University of Tennessee, Knoxville, TN, 1989. [5] H.V. Shurmer, An electronic nose: a sensitive and discriminating substitute for a mammalian olfactory system, IEE Proceedings, 137 G (1990) 197-204. [6] T. Leen, M. Webb and Steve Rehfuss, Encoding and Classification in a Model of Olfactory Cortex. In: Proceeding of the 1991 International Joint Conference on Neural Networks (IJCNN'91). ISBN: 0-7803-0164-1. IEEE Press, Piscataway, NJ, 1991, Vol. 2, pp. 553-559. [7] P.E. Keller, et al., Transmission of Olfactory Information for Telemedicine, Interactive Technology and the New Paradigm for Healthcare, K. Morgan, R.M. Satava, H.B. Sieburg, R. Matteus, and J.P. Christensen, (ed.s), IOS Press, Amsterdam, The Netherlands, 1995, 168-172. [8] P. Érdi and G. Barna, Neurodynamic Approach to Odor Processing, Proceeding of the International Joint Conference on Neural Networks (IJCNN'91), ISBN: 0-7803-0164-1, IEEE Press, NJ, 2, 1991, 653-656. [9] T. Nakamoto, et al., Gas/Odor Identification by Semiconductor Gas Sensor Array and an Analog Artificial Neural Network Circuit, Proc. International Conference on Microelectronics, Bandung, Indonesia, 4, 1992, 1-9. [10] T. Nakamoto et al. Odor recorder using active odor sensing system, Sensors& Actuators, B 76, 2001, 465-469. [11] J.W. Gardner, Electronic nose developments at Warwick, Electronic Engineering, 62, (1990) 15. [12] P.E. Keller, et al., Three Neural Network Based Sensor Systems for Environmental Monitoring, Proceedings of IEEE Electro 94 Conference, Boston, MA, 1994, 377-382. [13] P.E. Keller, et al., Electronic Noses and Their Applications, Proceedings of the World Congress on Neural Networks'96, Mahwah, NJ, USA, 1996, 928-931.

[14] R.M. Stuetz, G. Engin, R.A. Fenner, Sewage odor measurements using a sensory panel and an electronic nose, Water SCI Technology, 38 (3): 1998, 331-335. [15] W. Bourgeois and R.M. Stuetz, Measuring wastewater quality using a sensor array: prospects for real-time monitoring, Water SCI Technology, 41, (12): 2000, 107-112. [16] T. T. Mottram, et al., J. M. Techniques to Allow the Detection of Oestrus in Dairy Cows with an Electronic Nose, in Electronic Nose and Olfaction 2000, Gardner, J. W.; Persaud, K. C., editors; IOP [17] R.E. Baby, et al., 2000, Electronic nose: a useful tool for monitoring environmental contamination, Sensors and Actuators B 69 (3): 214-218. [18] T. Nakamoto and H. Hiramatsu, Study of odor recorder for dynamical change of odor using QCM sensors and neural network, Sensors and Actuators, B 85, 2002, 263-269. [19] B. Karlik and Y. Bastaki, Real Time Monitoring Odor Sensing System Using OMX-GR Sensor and Neural Network, WSEAS Transactions on Electronics, 1/2, 2004, 337-342. [20] P. Boilot, et al., Detection of Bacteria Causing Eye Infections using a Neural Network Based Electronic Nose System, in Electronic Nose and Olfaction 2000, 189-196. [21] A.K. Pavlou and A.P Turner, Sniffing Out the Truth: Clinical Diagnosis Using the Electronic Nose, Clin. Chem. Lab. Med. 2000, 38(2): 99-112. [22] A.K. Pavlou, et al., Use of an Electronic Nose System for Diagnoses of Urinary Tract Infections, Biosens. Bioelectron. 2002, 17(10), 893-899. [23] D.E. Rumelhart, G.E. Hinton and R.J. Williams, Learning internal representations by error propagation. In D.E. Rumelhart and J.L. McClelland (Eds.) - Parallel Distributed Processing: Explorations in the Microstructures of Cognition, Vol. 1, MA, 1986. [24] Y. Ozbay and B. Karlik, A Fast Training Back-Propagation Algorithm on Windows, Proceedings of the Third International Symposium on Mathematical & Computational Applications, pp. 204-210, 4-6 September, 2002, Konya, Turkey

APPENDICES

Appendix A

OMX-GR SENSOR Main parts name and function

No.

Display

Function Odor strength is indicated in numeric values (0 to 999). If more than 999 is indicated, 999 flashes. Odor is classified with two digit numeric values (0 to 89). "- -" is indicated for small amount of odor or when odor cannot be classified. Usually indicates elapsed time as hh:mm:ss format after the measurement starts. During the "Memory Sampling" period, the remaining time of data storage flashes. During zero-adjustment mode, "0" flashes at the "Second" position. Indicates the partition where data is being saved during "Memory Sampling" mode. When "Memory Sampling" is not selected, it indicates the partition where the data will be saved at the next Memory Sampling. Indicates the possible sampling period; RT, 1, 2, 5, 10, 20, 60, 120, or 300. Data saving period (seconds) can be selected. Regardless of this selection, strength and classification Displayed on the body are updated at real time (every second). Flashing Lit

: indicates warm-up period.

: indicates warm-up completion.

After the Odor Meter is powered on, the meter keeps warming-up for two minutes during which measurements cannot be made. Place meter in as clean a place as possible during "Power-on".

Lit

indicates the Peak-Hold is operating.

Battery life is indicated as number of Sufficient Continuous operation for 3 to 4 hours possible. Continuous operation for 1 to 2 hours possible. Battery replacement is recommended. No measurement possible. Replace the batteries immediately.

Appendix B

ANN OUTPUTS

Eta: 0.900000 Alfa: 0.700000 Iteration: 5000 MLP: 26 20 10 0.908678

0.131438

0.000013

0.000004

0.000180

0.000000

0.000000

0.102616

0.095265

0.000000

0.105691

0.743309

0.401860

0.112553

0.001612

0.000000

0.000000

0.082963

0.114329

0.000000

0.020229

0.319411

0.677711

0.308701

0.006535

0.000000

0.000000

0.087175

0.106219

0.000000

0.001120

0.001305

0.404036

0.640509

0.098877

0.000001

0.000000

0.104885

0.080110

0.000000

0.000021

0.000000

0.030940

0.338367

0.715363

0.098159

0.012207

0.143242

0.055520

0.000000

0.000003

0.000000

0.002473

0.021889

0.153236

0.695255

0.296167

0.082191

0.072035

0.000000

0.000002

0.000000

0.000386

0.000506

0.001015

0.237652

0.628346

0.114871

0.121974

0.000000

0.000001

0.000000

0.000018

0.000001

0.000000

0.002408

0.143634

0.676877

0.092919

0.001861

0.000001

0.000000

0.000005

0.000000

0.000000

0.000299

0.172781

0.104486

0.693782

0.265865

0.000001

0.000000

0.000004

0.000000

0.000000

0.000150

0.152946

0.013725

0.229359

0.669087

Appendix C

TEST DATA 0.64558 0.64556 0.64543 0.64543 0.64541 0.64543 0.64543 0.64541 0.64541 0.64541 0.64541 0.64541 0.64541 0.64528 0.64528 0.63529 0.63529 0.64541 0.64541 0.64541 0.64541 0.64541 0.64541 0.64541 0.64528 0.64528

0.60502 0.60489 0.61485 0.60487 0.61485 0.61513 0.60515 0.605 0.60528 0.60528 0.605

0.605

0.60487 0.605

0.61526 0.60528

0.61483 0.61485 0.60472 0.60472 0.60461 0.6147 0.60487 0.60487 0.61485 0.60487

0.5704 0.5704 0.5704 0.58041 0.58041 0.58041 0.58041 0.57042 0.57042 0.58043 0.58043 0.58043 0.58043 0.58043 0.57042 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041 0.58041

0.71083 0.72087 0.71091 0.71095 0.71099 0.72103 0.71099 0.72099 0.72099 0.73098 0.72094 0.73094 0.72091 0.72091 0.72091 0.72095 0.72095 0.72095 0.72095 0.71095 0.71095 0.72103 0.71103 0.71103 0.71103 0.72103

0.68209 0.68209 0.68216 0.68216 0.68223 0.68223 0.68223 0.68223 0.6823 0.6823 0.6823 0.6823 0.6823 0.6823 0.68223 0.68223 0.68223 0.68223 0.6823 0.67224 0.67217 0.67217 0.68223 0.68223 0.68223 0.68223

0.71046 0.71046 0.71046 0.71046 0.71046 0.71046 0.71046 0.71046 0.71046 0.70046 0.69046 0.70046 0.70046 0.69046 0.69046 0.70048 0.70048 0.70048 0.70049 0.70049 0.70049 0.70048 0.70049 0.70048 0.70046 0.70046

0.61096 0.61096 0.60093 0.60092 0.60093 0.60093 0.61096 0.61096 0.61096 0.61096 0.60092 0.60092 0.60092 0.60092 0.61092 0.61092 0.60089 0.60089 0.60089 0.60089 0.60089 0.60092 0.60089 0.60089 0.60089 0.61088

0.68132 0.69137 0.69137 0.68137 0.68137 0.69142 0.69142 0.68137 0.68137 0.68137 0.69142 0.69142 0.69142 0.69142 0.69142 0.68142 0.68142 0.69142 0.69142 0.69142 0.69142 0.69142 0.69142 0.68137 0.69137 0.69137

0.7159 0.71575 0.71574 0.7156 0.7156 0.71559 0.72573 0.71559 0.71559 0.71559 0.71559 0.72588 0.72588 0.72587 0.72573 0.72572 0.72571 0.72557 0.72557 0.72556 0.72556 0.72557 0.72556 0.72542 0.72542 0.72556

0.74631 0.74649 0.7465 0.7465 0.7465 0.74632 0.74632 0.74632 0.73633 0.7365 0.7365 0.74667 0.74667 0.7365 0.73633 0.73617 0.73617 0.73617 0.73617 0.73585 0.73584 0.736

0.73601 0.73601 0.73585 0.73585

0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.61999 0.62999 0.62999 0.62999 0.62999 0.62999 0.61999 0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.62999 0.61999 0.62999 0.62999

0.46014 0.46014 0.46014 0.46014 0.46014 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015 0.46015

0.82394 0.82382 0.82383 0.82372 0.82382 0.82372 0.82416 0.8244 0.82428 0.82416 0.82382 0.82372 0.82383 0.82394 0.82405 0.82394 0.82382 0.82383 0.82383 0.82383 0.82394 0.82394 0.82394 0.82394 0.82383 0.83382

0.78999 0.78999 0.78999 0.78999 0.79999 0.78999 0.78999 0.78999 0.79999 0.78999 0.78999 0.78999 0.78999 0.79999 0.79999 0.78999 0.78999 0.79999 0.79999 0.78999 0.79999 0.79999 0.79999 0.79999 0.79999 0.79999

0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056 0.85056

0.71999 0.71999 0.71999 0.70999 0.71999 0.71999 0.71999 0.71999 0.71999 0.71999 0.71999 0.71999 0.71999 0.70999 0.70999 0.70999 0.71999 0.71999 0.71999 0.70999 0.71999 0.71999 0.71999 0.70999 0.70999 0.70999

0.66805 0.65808 0.65808 0.66805 0.66805 0.65785 0.65785 0.65785 0.65785 0.66782 0.65765 0.66763 0.6676 0.6676 0.6676 0.66739 0.66722 0.66722 0.66722 0.67736 0.66719 0.66719 0.66701 0.66701 0.67717 0.66701

0.63076 0.63072 0.64072 0.64068 0.63065 0.64065 0.65064 0.64061 0.65061 0.63062 0.63062 0.63062 0.64065 0.63058 0.65061 0.65061 0.65058 0.65057 0.65055 0.66054 0.66054 0.65055 0.65055 0.64055 0.63052 0.64052

0.67197 0.67197 0.67204 0.6721 0.67204 0.67204 0.6721 0.6721 0.6721 0.6721 0.67204 0.66198 0.67197 0.67197 0.66198 0.67203 0.67203 0.66198 0.66198 0.66204 0.67211 0.66204 0.67217 0.66211 0.66212 0.66211

0.65151 0.66151 0.66151 0.66151 0.65146 0.66151 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.65146 0.66145 0.66145 0.66145 0.66145 0.66145 0.66145 0.66145 0.66145