CONTENTS Plenary Talk - 1 Progress in Wireless Sensor Design and Networking, David V. Thiel {Centre for Wireless Monitoring and Applications, Griffith University, Nathan Qld 4111 Australia}
1
Session 1. Power Engineering (PE-1) 001PE
005PE
006PE
009PE
015PE
Building Integrated Photovoltaic (BIPV): A Case Study, Fendy Santoso1, Hoga Saragih2 {1Department of Electrical and Computer Systems Engineering, Monas University, VIC 3800, Melbourne, Australia; 2Department of Electrical Engineering, Faculty of Engineering, University of Indonesia, Indonesia}
4
Dynamic Contact Resistance of Tap Changer, Indera Arifianto1, Bambang Cahyono2 {1PT.PLN (Persero) P3B Java Bali, West Java Region, Indonesia; 2 PT.PLN (Persero) P3B Java Bali, Jakarta and Banten Region, Indonesia}
10
The Simulation of Impulse Current Generator, Haryono, T1, Sirait, K.T2, Tumiran1, Hamzah Berahim1, {1Electrical Engineering Department Gadjah Mada University, Yogyakarta, Indonesia; 2School of Electrical Engineering and Information, Bandung, Indonesia}
13
Incremental Loading Analysis on Critical Buses Using Two-Port Network in 500 kV Java System, Herbert Innah1, Yosef Lefaan1, Matius Sau2 {1Electrical Department, Cenderawasih University Jayapura, Indonesia; 2Electrical Department, Paulus Indonesian Christian University Makassar, Indonesia}
17
Short-Term Load Forecasting Using Extreme Learning Machine, Hasmina Tari Mokui, Bunyamin {Faculty of Engineering, Haluoleo University}
21
Session 2. Telecommunication (TC-1) 001TC
003TC
004TC
005TC
vi
Portfolio Model for Cellular Telecommunication Industry in Indonesia Using Equally Weighted Method, Hoklie1, Syarif Hidayat1, Ary Syahriar2 {1Industrial Engineering Department, University Al-Azhar Indonesia; 2Electrical Engineering Department, University Al-Azhar Indonesia}
25
Multi-target SSVEP-based BCI using Multichannel SSVEP Detection, Indar Sugiarto {Department of Electrical Engineering, Petra Christian University, Surabaya, Indonesia}
29
Adaptive Modulation and Coding Performance on UWB OFDM System, Budi Prasetya, Saleh Dwi Mardiyanto, Gema Aji Morgana {Electrical Engineering Department, Institut Teknologi Telkom, Bandung, Indonesia}
34
Performance Assessment of VoIP over WiMAX using IPv4 and Ipv6: The Effect of Number of Users, Adi Nariswara, Rendy Munadi, Wiseto Agung {Department
007TC
of Magister Telecommunication, Institut Teknologi Telkom, West Java, Indonesia}
40
The Development of Video on Demand (VoD): Interactive Web Base System using TreeMap Information Visualization Representation, Muhammad Haziq Lim Abdullah, Nuridawati Mustafa, Norazlin Mohammed, Ezar Eziardy Zainudin {Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka, Locked Bag 1200, Hang Tuah Jaya, 75450 Ayer Keroh, Melaka}
43
Session 3. Computer Engineering & Informatics (CEI-1) 001CEI
003CEI
004CEI
005CEI
007CEI
A Vision-based Control Strategy Using a Low-end USB Camera for Gantry Crane System, Ricki Y. Setiawan, Novy N.R.A. Mokobombang {HAN University of Applied Sciences, HAN Kenniscentrum Economie, Techniek en Informatica, Ruitenberglaan 26, Arnhem, The Netherlands}
46
Runge-Kutta – Split-Step Fourier Method For Solving Generalized Nonlinear Schrödinger Equation in Nonlinear Fiber Optics, Endra {Department of Computer Engineering, University of Bina Nusantara, Jakarta, Indonesia}
53
Image Processing Algorithm in Machine Vision Approach for Industrial Inspection, Habibullah Akbar, Anton Satria Prabuwono, Zeratul Izzah Mohd. Yusoh, Zulkifli Tahir {Computer Vision and Robotics (CoVisBot) Lab., Faculty of Information & Communication Technology, Technical University of Malaysia Melaka}
58
Developing Multimedia Content for Diabetes Education (in English and Malay Language): A first Prototype in Developing Content to Educate, Motivate and Monitor (Public and Diabetic), Jasni M. Zain, Abbas Saliimi Lokman {Faculty of Computer Systems & Software Engineering, Universiti Malaysia Pahang, Lebuhraya Tun Razak, 26300, Kuantan, Pahang, Malaysia}
63
Design of Neural Network-based Intellegent Classroom System: A Preliminary Research, Dena Priyanto1, Arwin Datumaya Wahyudi Sumari12, Eko Patra Teguh Wibowo1 {1Department of Electronic, Indonesia Air Force Academy; 2 School of Electrical Engineering and Informatics, Bandung Institute of Technology, Indonesia}
67
Session 4. Control Engineering (CE-1) 002CE
003CE
Ziegler-Nichols Based PID Control Tuning for a pH Process, Fendy Santoso1, Hoga Saragih2 {1Department of Electrical and Computer Systems Engineering, Monash University, VIC 3800, Melbourne, Australia; 2Department of Electrical Engineering, Faculty Of Engineering, University of Indonesia, Indonesia}
72
Limited Speech Recognition for Controlling Movement of Mobile Robot Implemented on ATmega162 Microcontroller, Thiang, Dhanny Wijaya {Electrical Engineering Department, Petra Christian University, Surabaya, Indonesia}
80
vii
Proceedings of The 1st Makassar International Conference on Electrical Engineering and Informatics Hasanuddin University, Makassar, Indonesia November 13-14, 2008
007CEI
Design of Neural Network-based Intelligent Classroom System: A Preliminary Research Dena Priyanto#1, Arwin Datumaya Wahyudi Sumari#*2, Eko Patra Teguh Wibowo#3 #
1
*
Department of Electronics, Indonesian Air Force Academy Jl. Laksda Adisutjipto, Yogyakarta – 55002, INDONESIA
[email protected],
[email protected]
School of Electrical Engineering and Informatics, Bandung Institute of Technology Labtek VIII, Jl. Ganesha 10, Bandung – 40132, INDONESIA 2
[email protected],
[email protected]
Abstract— Intelligent rooms become a need in the information technology era especially for people who want their office or workplace equipment is ready to be used when they arrive there. In this paper we propose an Intelligent Classroom System (ICS) that a system which is able to prepare the classroom equipment intelligently and remotely before the lecture is started. The ICS uses a Back Propagation-type Neural Network (BPNN) as the interface between the control signal from Dual Tone Multi Frequency (DTMF) line and the AT89C51 microcontroller that controls the classroom equipment via a set of relays. One of the problems in transmitting signal controls remotely is there maybe some errors when the signals are received at the microcontroller. These errors may be caused by some noises in the transmitted signals. For this reason, the BPNN is aimed to minimize errors of the transmitted signals before being used by microcontroller to drive the equipment. Some simulation results will be presented to show that the system can be implemented in the real life. Keywords: DTMF, intelligent, ICS, microcontroller, neural network, remote control
I. INTRODUCTION The fast growing of the Information and Communication Technology (ICT) affects the development in many engineering fields. This growing technology brings many benefits to the human kind. One of these benefits is we can create intelligent buildings, smart houses, or anything to ease our common everyday workloads. As lecturers or students, we need a classroom that is well prepared before the lecture is started. Usually, before the lecture is begun, an office person prepares the all classroom equipment such as laptop, LCD projector, room lighting, air conditioner, etc. If he is late for some reasons, we have to prepare it ourselves and it will take a time so the lecture will be delayed. In this paper we propose and design an Intelligent Classroom System (ICS) to cope with the common problem as mentioned before. The ICS makes use of Dual Tone Multi Frequency (DTMF) signals as the system’s control inputs. DTMF signals are used by AT89C51 microcontroller to drive
a set of relays that are connected to the classroom equipment. The kind of equipment will be activated depends on the keypad number that is pressed. In order to handle imperfect control input signals that probably occur during signals transmission, we use Back Propagation-type Neural Network (BPNN) to compensate the errors at minimum level. The BPNN is installed between DTMF and AT89C51 microcontroller. Its primary job is to recover the imperfect control input signals from DTMF before being fed to AT89C51. The system’s intelligence is presented by the capability of the BPNN in handling imperfect inputs automatically without human intervention. The paper will be begun with Section I which covers the paper’s background and a little introduction of the ICS. In Section II some background theories will be described briefly and the system design will be explained in Section III as well as some simulation results. Section IV ends the paper with some concluding remarks. II. FOUNDATION THEORIES In this section, some foundation theories such as telephone controlling and NN as well as other related matters will be described briefly. A. Telephone Controlling
Fig. 1 Schema of DTMF telephone system [5]
ISBN: 978-979-18765-0-6
67
Proceedings of The 1st Makassar International Conference on Electrical Engineering and Informatics Hasanuddin University, Makassar, Indonesia November 13-14, 2008
007CEI
The equipment remote controlling can be carried out by using a DTMF telephone control system via telephone line as the transmission medium. DTMF detector converts the analogue signals received from telephone keypad to digital signals that then will be used by microcontroller to drive the equipment via a set of relays. The schema of a DTMF telephone system is depicted in Fig. 1. B. Neural Networks Artificial Neural Networks (ANN) or just Neural Networks (NN) is an information-processing system that has certain performance characteristics in common with biological NN [4]. NN is generalization of mathematical models of human cognition based on the assumptions that [4]: • information processing occurs at many simple elements called neuron, • signals are passed between neurons over connection links, • each connection link has an associated connection weight which multiplies the signals transmitted, • each neuron applies an activation function which is usually non linier, to its net input to determine its output signals.
C. Back Propagation Neural Networks The BPNN was developed to cope with the limitations of single-layer NN. The BPNN actually is a feedforward NN that is trained by backpropagation which means the signals is propagated in reverse direction. The primary aim of NN training is to train the NN to achieve a balance between the ability to respond correctly to the input patterns that are used for training or memorization, and the ability to give reasonable responses to input that is similar that used in training or generalization. In training the NN with backpropagation mechanism, there will be three steps, namely: • feedforwarding the input training patterns to the NN input layer, • calculating the NN outputs and backpropagating the associated error, • adjusting the connection weights to minimize the error. After passing the training session, the NN uses the final connection weights when recognizing the input patterns given to it. As commonly in a multilayer NN, BPNN has three layers namely input layer, hidden layer, and output layer. The number of input and output layers is depended on the input pattern and the output’s target. The number of hidden layer is depended on particular applications, but commonly one hidden layer is sufficient for many applications. The architecture of BPNN is depicted in Fig. 3.
Fig. 2 The architecture of the NN model [8].
In NN model, neuron takes a set of inputs, xm , along with a set of connection or link or synaptic which are characterized by weights, wkm . The summing junction, ∑, sums up the input signals that are amplified by the connection weights. The activation function, ϕ ( . ) , limits the net outputs in allowable values. The architecture of the NN model is depicted in Fig. 2, while the general mathematical equations for neural information processing are given in Eq. 1 for inputs summing process to obtain v k and Eq. 2 for producing the NN output,
yk .
1) Training Algorithm: Refer to [4] for detailed BPNN training algorithm.
m
vk = ∑ wkj x j
(1)
j =0
yk = ϕ ( v k )
Fig. 3 The architecture of the BPNN model
(2)
2) Activation Function: There are four common NN activation functions namely: • identity function which produces binary output (0 or 1), • binary step function with threshold θ ,
ISBN: 978-979-18765-0-6
68
Proceedings of The 1st Makassar International Conference on Electrical Engineering and Informatics Hasanuddin University, Makassar, Indonesia November 13-14, 2008
binary sigmoid (logistics sigmoid), bipolar sigmoid, hyperbolic tangent. The sigmoid function and hyperbolic tangent are the most common activation function for training NN with backpropagation mechanism. • • •
D. Related Researches Some researches have been done in using the DTMF to control electrical and electronics equipment such as done by [3]. The use of microcontroller AT89C51 for controlling the electrical equipment has been done by [5]. The application of BPNN in controlling a DC motor has also been already simulated in [6].
007CEI
TABLE I RELATION OF KEYPAD-RELAYS AND ITS BINARY INPUT CONVERSION [2]
No
1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
III. DESIGNING THE ICS This section consist of three parts, namely ICS block diagram and building and training the BPNN. A. ICS Block Diagram The ICS block diagram is obtained by adding BPNN block between DTMF and microcontroller AT89C51 as presented in Fig 4.
Fig. 4 ICS block diagram
1) Inputs Definition: For this preliminary research, we design the system to drive four relays that are connected to one of four classroom equipment. The system is connected to the telephone line and only accepts keypad button number between ‘0’ and ‘9’. Each keypad button is assigned a particular instruction. The relays are driven by the AT89C51 microcontroller after receiving signals from BPNN. The input to BPNN is DTMF 4-bit digital signal. The relation between keypad numbers and possible combinations of relays is presented in Table I, while the relation between relays and the classroom equipment is presented in Table II.
Keypad Button Number 1 2 3 4 5 6 7 8 9 0
Combinations of Relays
1,2,3,4 2,3,4 1,3,4 1,2,3 3,4 2,3 1,3 1,2 3 -
Conversion to Binary Input (from Keypad Button Number) 0001 0010 0011 1000 1001 1010 1011 1000 1001 1010
TABLE II RELATION OF RELAYS-CLASSROOM EQUIPMENT
No
Relay Number
1. 2. 3. 4.
1 2 3 4
Equipment
Air conditioner Room lighting Laptop LCD projector
Active Binary Code 1 1 1 1
Inactive Binary Code 0 0 0 0
2) Output Definition: The output of the ICS is the combination of active-inactive relays signals that are used to drive the classroom equipment. Because the BPNN needs target patterns to be compared with its output patterns, we use the combinations of active-inactive relays as the target patterns. To obtain these patterns, we converts the relay combinations into binary patterns where the active relays are assigned with binary ‘1’ while the others with binary ‘0’. This system is used high logic to active-inactive switch line system. The relation between binary inputs with binary target patterns is presented in Table III. TABLE III RELATION OF KEYPAD-RELAYS AND ITS BINARY TARGET CONVERSION
No
1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Binary Input
0001 0010 0011 1000 1001 1010 1011 1000 1001 1010
Combinations of Relays
1,2,3,4 2,3,4 1,3,4 1,2,3 3,4 2,3 1,3 1,2 3 -
Conversion of Binary Target (from Combinations of Relays) 1111 0111 1011 1110 0011 0110 1010 1100 0010 0000
ISBN: 978-979-18765-0-6
69
Proceedings of The 1st Makassar International Conference on Electrical Engineering and Informatics Hasanuddin University, Makassar, Indonesia November 13-14, 2008
007CEI
B. Training the BPNN
1 0 1 1 0 T = 0 1 1 0 0
Fig. 5 BPNN architecture for ICS
Before training the NN, we have to set up the number of neurons in each layer. For the ICS, there will be 4 neurons in the input layer, 2 neurons in the hidden layer, and 4 neurons in the output layer. For training the BPNN, we use the source codes taken from [1]. All BPNN learning parameters such as goal, learning rate, and maximum performance, use the default value except for epoch. The activation function for all layers is bipolar sigmoid function (tansig). The BPNN architecture is depicted in Fig. 5.
TABLE IV TARGET PATTERNS
C
0 0 1 0 1 0 0 1 1 1 0 0 1 0 1 1 1 0 1 1 1 0 0 0 0 0 1 0 1 0
(4)
The NN never converges to the default Mean Squared Error (MSE), 0.001. After 50.000 epochs, the MSE achieved by the NN is 0.0206986 which is the best result that we can get. The target patterns and the BPNN output patterns after learning phase is presented in Table IV and Table V while the simulation results are presented in Fig. 6 and Fig. 7.
C. Simulation Results In training phase, the input patterns and the target patterns are organized in matrices form. The input pattern and target pattern matrices have dimension of 4 x 10 as presented in Eq. 3 and Eq. 4 successively.
0 0 0 0 0 X = 0 0 1 1 1
1 1 1 1 1 1 0 1 1 1 1 0 0 1 1 1 1 0 0 1 0 1 0 0 0 1 0 0 0 0
1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Y0 1 0 1 1 0 0 1 1 0 0
Target Patterns Y1 Y2 1 1 1 1 0 1 1 1 0 1 1 1 0 1 1 0 0 1 0 0
Y3 1 1 1 0 1 0 0 0 0 0
TABLE V BPNN OUTPUT PATTERNS
(3) No 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
Y0 1.0000 0.0002 0.9998 1.0000 0.0003 0.0054 1.0000 0.8808 0.0680 0.0000
BPNN Output Patterns Y1 Y2 Y3 0.9999 1.0000 1.0000 1.0000 0.9999 0.9999 0.0065 0.9926 0.9924 1.0000 0.9246 0.1795 -0.0200 0.9966 0.9995 1.0000 0.8834 0.1166 0.0403 0.8731 0.2212 0.9015 0.2336 -0.1146 0.0443 0.4040 0.0492 0.0000 0.0000 0.0000
ISBN: 978-979-18765-0-6
70
Proceedings of The 1st Makassar International Conference on Electrical Engineering and Informatics Hasanuddin University, Makassar, Indonesia November 13-14, 2008
007CEI
worst result is output pattern for target output “0010” (9) where the result is 0.0680 for ‘0’, 0.0443 for ‘0’, 0.4040 for ‘1’, and 0.0492 for ‘0’ or resembles pattern “0000” are presented in fig. 8. So, percentage the worst result is 0.4040/1*100% = 40.40%. But as we can observe from Table V, in general, the BPNN shows a good result because from 10 input patterns only 1 pattern that is failed to be recognized or 90% success[7].
Performance is 0.0206986, Goal is 0.001
1
10
0
Training-Blue Goal-Black
10
-1
10
-2
10
-3
10
-4
10
0
0.5
1
1.5
2 2.5 3 50000 Epochs
3.5
4
4.5
5 4
x 10
Fig. 6 BPNN performance after 50.000 epochs
Comparison between Target (o) and Output Layer(*) Target or Output
1
0.5
0
-0.5
0
0.1
0.2
0.3
0.4
0.5 0.6 0.7 0.8 First Input Comparison between Target (o) and Output Layer(*)
0.9
1
Target or Output
1
0.5
ACKNOWLEDGMENT Writers would like to thank and give a big appreciation to the Governor of the Indonesian Air Force Academy for supporting the research and hopefully further researches in this field as well as other supporting staffs that cannot be mentioned here.
0
-0.5
REFERENCES [1] 0
0.1
0.2
0.3
0.4 0.5 0.6 Second Input
0.7
0.8
0.9
1
[2]
Fig. 7 Map of target patterns (o) vs BPNN output patterns (*)
[3]
Comparison Between Output BPNN and Target
[4] [5]
1.2 1
0.9998 1 0.8808
1 0.9999 0.9015
1 0.9999 0.9966 0.9926 0.9246 0.8834 0.8731
0.9995 1 0.9999 0.9924
[6]
Series4
0.6
Series5
0.4
0.404
0.2
0.2336
0
Series1 Series2 Series3
0.8 Output BPNN
IV. CONCLUDING REMARKS We have presented ICS design along with its simulation results. By observing the BPNN results, further studies must be worked out to bring the BPNN to its most optimal performance. There are some options that can be done such as modifying the activation function in each layer and do some conversion to the input patterns so the bit difference between patterns is only 1 bit. The results of BPNN simulation in recognizing the target patterns is satisfying enough at this step, and this research can be a stepping-stone for promising works in the future in this field.
0.068 0.0054 0.0003 0 0.0002
0.0443 0.0403 0.0065 0 -0.02
0
Series6 Series7 0.2212 0.1795 0.1166 0.0492 0
Series8 Series9
[7]
Series10
-0.1146 -0.2 Target
Fig. 8. Map of target patterns (X axis) vs BPNN output patterns (Y axis)
[8] [9] [10]
S. Kusumadewi, Building Artificial Neural Networks using MATLAB & EXCEL LINK, Grha Ilmu, Yogyakarta, Indonesia, 2004 (in Indonesian). Application of the MT8888 Integrated DTMF Receiver with Micro Interface, Mitel, 1997. S. Parulian, “Electronic Equipment ON-OFF Controller using DTMF Signal on Telephone Set”, Elektron Tabloid, No. 52, XXV, pp. 49-54, Bandung, Indonesia, 2003 (in Indonesian). L. Fausset, Fundamentals of Neural Networks: Architectures, Algorithms, and Applications, Prentince-Hall, USA, 1994. M. Syaryadhi et.al., “Electrical Lamp Controlling System via Telephone Line based on AT89C51 Microcontroller”, Journal of Rekayasa Elektronika, Vol. V, No. 2, University of Syiah Kuala, NAD, Indonesia, 2006 (in Indonesian). R. Wiryadinata et.al, “Simulation of Back Propagation Artificial Neural Networks as DC Motor Speed Controller”, Proc. of National Seminar on Information Technology Application, Islamic University of Indonesia, 2005, Yogyakarta, Indonesia, pp. C.29-35 (in Indonesian). D. Priyanto, “Simulation & Implementation Concept: Design of Neural Network Backpropagation Methode-based Intelligent Classroom System”, Final Project to achieve 2nd Lieutnenant of the Indonesian Air Force, Department of Electronics, Indonesian Air Force Academy, Yogyakarta, Indonesia, October 2008 (in Indonesian). S.J. Russell, “Artificial Intelligence: A Modern Approach”, PrenticeHall, USA, 2003. Microcontroller to Telephone Line. [Online]. Available: http://www.delta-electronic.com. 8-bit Microcontroller with 4K Bytes In-System Programmable Flash AT89S51. [Online]. Available: http://www.atmel.com.
The best results are BPNN output patterns for target patterns “1111” (1) and “0000” (10) or 100% correct. The
ISBN: 978-979-18765-0-6
71
