(IJCSIS) International Journal of Computer Science and Information Security

Vol. 9, No. 5, May 2011
A Novel Model for Synchronization and Positioning
by using Neural Networks

Hossein Ghayoumi Zadeh*, Siamak Janianpour and Javad Haddadnia
Department of Electrical Engineering
Sabzevar Tarbiat Moallem University
Sabzevar, Khorasan Razavi, Iran
Email: h.ghayoumizadeh@gmail.com*

Abstract— In this paper by using a Low Noise Amplifier (LNA), a
synchronization and positioning system is designed. Parameters
that cause the system to be far from ideal condition such as
S-Parameters, Noise Figure, IIP3, and Gain are considered that
is one of the advantages of this system. In this stage this process is
a little slow so by adding the neural network to the system the
speed of synchronization is increased. By using the neural
network the time needed to calculate the time difference of
arrival (TDOA) is significantly decreases.
Keywords-component; Global Positioning System, Low-noise
amplifiers, Neural networks.
The TDOA approximation has so many different
applications such as communication, electronic war and
medical engineering. Following some of these applications will
be discussed. One of the most important applications of TDOA
is in positioning of the transmitters. Nowadays radar and sonar
systems are widely used with many different military or
nonmilitary applications and their importance in security
problems are so that they are parts of the strategic system of
each country so to protect the radars; the usage of the passive
radar is become popular, increasingly[1]. Although in these
type of radars the basic principle of the radars are dominant but
the transmitter of the radar is omitted from the system and by
omitting the transmitter the receiver will become hidden from
the sight of the enemies. The place of transmitter is one of the
most important parameters that assign the duties of the radar.
Another application of TDOA is in measurement and
controlling the coolant current of the atomic reactors. Also it
can be used to localize the position of brain that controls the
simultaneity of the activities in epileptic patients. Further it is
used in global positioning system (GPS), recently. Contrary to
other methods, this system will not affect the normal operation
of the satellites, because the time delay is calculated passively.
Also there is no need to carry extra hardware in spaceships and
this will reduce the cost of this procedure.
By calculating the received uplink signal TDOA using three
or four satellite on earth orbit, the position of the transmitters
can be localized. When the transmitter is located on the earth,
three satellites and when the elevation of the transmitter is not
known, four satellites are needed to localize the position of the
transmitters. So in the passive systems, approximation of the
delay time between the signals receive time from two different
sensors plays a significant role in measurement of the distance
and direction of the transmitters. When a signal is emitted into
the environment from transmitter, it spreads with a specific
speed, so two receivers with different distance from the
transmitter, sense the signal with a time delay. If S(t) is the
emitted signal from transmitter and assume that there is just
one way for signal transmission, then the signals first will
receive to the nearest receiver and with a delay to the next
receiver, the delay time is shown with D. So the goal is to
measure this delay and approximation of this delay has a
significant role in synchronization and positioning process [2].
The controll system is shown in fig. 1. In this circuit the
LNA model is used in S-parameters block. In this block the
values of S-parameters, Noise Figure and IIP3 are changeable.
In this circuit the Gaussian Signal is used as input
(Fig. 2). Also the white noise is applied to the signal and
considered as a non ideal factor.

Figure 1. Controlling system designed for Synchronization
48 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 5, May 2011

Figure 2. The Gaussian Signal that is applied to the control
circuit as input
The structure of the synchronization is so that the input
delay of T1 is considered by the integer delay block. After the
amplification stage, the changes of the S-parameter are exerted
to the input signal that simulates the LNA stage, and it is
transferred to the output. The second input in this simulation
that is also the Gaussian signal, is multiply to the first signal
with a delay and it is transferred to the output. The point is that
the second output also has delay equal to T2 that is shown in
the “integer delay 2” block. Now by using a feedback from the
output to the second input, the delay is changeable so that the
both inputs become concurrent.
The easiest and most effective way for synchronization of
the both input signals is to multiply the feedback output with
the first input, and simultaneously check the output until the
output becomes maximal. Now the delay can be recorded and
stored. When the both inputs become synchronize the output
will become maximal.
In this paper the model of an inductorless low-noise
amplifier (LNA) is used. [3] This LNA is designed for ultra-
wideband (UWB) receivers and for microwave access,
covering the frequency range from 0.4 to 5.7 GHz using 0.18-
μm CMOS technology. Simulation results show that the
voltage gain reaches a peak of 19.6 dB in-band with an upper

Figure 3. LNA circuit that is used in the controlling circuit.
3-dB frequency of 5.7 GHz. The IIP3 is about -3 dBm and the
noise figure (NF) ranges from 3.06-3.8 dB over the band of
interest. Input reflection coefficient S11 is below -8.79dB for
the design. The LNA consumes 5.77 mW from a low supply
voltage of 1.8 V. A figure of merit is devised to compare the
proposed designs to recently published wideband CMOS
LNAs. The proposed topology achieves a lower NF than that of
the topology capacitive cross-coupling with inductors, with the
additional advantage of removing the bulky inductors. It is
shown that the LNA is designed without on-chip inductors that
its performance is comparable with inductor-based designs.
The LNA circuit that is used in this system is shown in fig. 3.
LNA properties such as gain, IIP3 and etc. that are used for
synchronization in this system are given in TABLE I.
Also an example of the test results is shown in

FOM No.of Coils Power(mW) IIP3 (Dbm) Gain (db) S
(dB) NF (dB) BW(GHz) Technology
3.68 0 5.77 -3 19.66 <-8.79 3.06 ~ 3.8 0.4 ~ 5.7 0.18μm
S22 S21 S12 S11 Frequency OIP3(dbm) Noise Figure (db) Gain Delay Gain1 Output
-8.64 -49.76 1.76 -9.32 3.5E6 inf 0 100 5 -10
49 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 5, May 2011
By evaluation of the test results, some points are
1. The delay time is respectively related to noise, noise
figure, initial delay and OIP3. The LNA gain will affect
the delay time but it has a little impact. Also the impact
of S-parameter is so small that it is neglect able.
2. By increasing the noise figure the delay time will
increase. Also in high value of the LNA gain its
increase will cause an increase in delay time, but noise
figure does not show such a behavior.
3. The effect of the OIP3 in comparison with S-parameter
is higher. (by changing the S-parameter values not a big
change is observed.
4. By increasing the OIP3 value the delay time is
But the problem in designing such a controlling system is
that about 45 second is needed to process such a big amount of
data and it is one of the disadvantages of this system, because
moreover to input delay time we will lose the time needed for
the calculation process of software that is very bad for
synchronization systems. So to reduce the calculation time the
neural network is used.
In this system a linear neural network is used. To minimize
the errors in these networks, the training process is done by
using squares least mean algorithm (Fig. 4).
The neural network that is used in this process must be so
fast and so accurate. First of all the goal is to synchronize the
signals with each other by using the neural network. To train
the neural network the previous Gaussian signal is used that is
shown in fig. 5.
The objective function of the neural network can be
achieved from multiplying of two Gaussian signals that is
shown in fig. 6.
Now many samples of Gaussian signal with different delay
time are applied to the neural network for training purpose. For
instance to test the designed neural network the signal that is
shown in fig. 7 is used. Also to test the robustness of the
network, samples that contain noise and have different delays
in comparison with original signals are used.

Figure 4. The structure of linear neural network used in this control system.

Figure 5. The initial signal that is used to train the neural network

Figure 6. The objective signal that is used to train the system.
An example of the signal that is used for test purpose is
shown in fig. 7. The noises and delays are obviously shown in
this figure.

Figure 7. The test signal that contains noise and delay
50 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 5, May 2011
Now the results achieved by using neural networks are
shown in fig. 8. The neural network tries to find the appropriate
delay time that leads to the objective function by using the
iterative methods. If you look at the fig. 8 carefully you will see
that some parts of the chart are drawn thicker and or bolded.
That is caused by increasing the accuracy of the neural network
and the number of iterations. In other words, in this method
two signals are studied by different random delays and the
delay that cause the output to reach the objective function is
In the proposed method the noise in the Gaussian signal,
has a small effect on calculation of the delay time that is clearly
shown in fig. 8.
By decreasing the accuracy of the network the segregation
of the signals with different delay time are shown more clearly
(fig. 9).

Figure 8. The output of the neural network.

Figure 9. The more detailed chart of the output signal.
In this method the speed of delay calculation in
significantly increased and the calculation time is decreased. It
is good to mention that the Gaussian signal enters the system
periodically and the system must be able to sweep the input
continuously. To consider the noise in this system, for
synchronization other techniques must be added to this system.
Now we are going to describe these techniques. Assume a
signal like the signal in fig. 10 is given to the system as input.
First an intact period of the signal must be given to the system
as the training sample.
There are so many methods for positioning systems based
on the calculation of the signal time-of-flight. One of these
methods is TDOA. In this method to calculate the position of
the TAG, the distance between the TAG and a node is
calculated and it is compared with the distance of the TAG
with another node. In this method the TAGs are just the
transmitter and the nodes are just receiver. In 2 dimensional
systems 4 TAGs and in 3 dimensional systems 5 TAGs are

Figure 10. Three period of signal with consideration the effect of noise.

Figure 11. Geometrical structure of a 3D TDOA.
51 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 5, May 2011
In this study the unknown position of the TAG is shown
with function E (Equation 1).
E=(x,y,z) (1)
And the receivers are defined as (Equation 2):
, P
, ..., P
, ..., P
= (x
, y
, z
), 0 ≤ m ≤ N (2)
Where N is the function dimension.
The distance between the each transmitter and the receiver
is defined as Rm and R0 is the distance between transmitter
and the false origin (it is assumed that one of the receivers is
located at the false origin). As it is shown in fig. 11 the
resultant time is calculated as follows (Equation 3):

m m
0 0 0t ÷ = (3)
In this equation,
0t is the time that the signal needs to
arrive to the m
receiver. The delay time that is considered in
this stage must refer to the calculated time in synchronization
stage. Now, the time duration (τ
) can be calculated by using
correlation function
( )
Substituting R
instead of
T 0 in above equation and after
some rearrangements, finally we will have a line equation with
constant coefficients (equation 4)[4,5].
0 = + + + x D x C x B x A
m m m m
That A
, B
, C
and D
in equation 4 are defined as

2 2
0t 0t
x x
÷ =

2 2
0t 0t
y y
÷ =

2 2
0t 0t
z z
÷ =

2 2 2
0t 0t
0t 0t
z z x z z x
m m m
m m
+ +
+ +
÷ ÷ =
It is very easy to calculate the three unknown parameters
(x,y,z) that are the coordinates position of the transmitter, by
solving the three equations with one of the different methods
such as singular value decomposition, numerical analysis or
etc. In this study the numerical analysis is used. TDOA is
mostly used in open areas for personal uses, air traffic control
(ATC) or military systems.
A sample of positioning process is shown in fig. 12.
Because of linearity of the equations, theoretically there
isn’t any error in this process but the practical problem, is the
signal transfer time calculation from the transmitter to the
receiver. In this paper the role of clock pulse generation and
synchronization is significant. To increase the accuracy of the
system, the clock pulse must be synchronized in every node. If
the generated clock pulse was in the order of 1ns, the best
achievable accuracy in the positioning system is about 30
centimeter that means we can find the position of the
transmitter with maximum error of 30 centimeter.

Figure 12. Positioning with TDOA method
[1] Fujiwara, R.; Mizugaki, K.; Nakagawa, T.; Maeda, D.; Miyazaki, M.; ,
"TOA/TDOA hybrid relative positioning system using UWB-IR," Radio
and Wireless Symposium, 2009. RWS '09. IEEE , pp.679-682, 18-22
Jan. 2009.
[2] Dawei, L.; , "Application of assisted TDOA positioning technology in
vehicle positioning and navigation," Mobile Technology, Applications
and Systems, 2005 2nd International Conference on , pp.5 pp.-5, 15-17
Nov. 2005.
[3] Ali Shirzad Nilsaz, Mohsen Khani Parashkoh, Hossain ghauomy-zadeh,
Zhuo, Zou, Majid Baghaei-Nejad, Li-Rong Zheng. “low power .18um
cmos ultra wideband inductor-less lna design for uwb receiver”.Asia
Pacific Conference on Circuits and Systems,2010.
[4] Kian Meng Tan; Choi Look Law; , "GPS and UWB Integration for
indoor positioning," Information, Communications & Signal Processing,
2007 6th International Conference on , pp.1-5, 10-13 Dec. 2007.
[5] Rabinowitz, M.; Spilker, J.J., Jr.; , "A new positioning system using
television synchronization signals," Broadcasting, IEEE Transactions on
, vol.51, no.1, pp. 51- 61, March 2005.
[6] Wenjuan Jia; Chunlan Yang; Guocheng Zhong; Mengying Zhou; Shuicai
Wu; , "Fetal ECG extraction based on adaptive linear neural network,"
Biomedical Engineering and Informatics (BMEI), 2010 3rd International
Conference on , vol.2, pp.899-902, 16-18 Oct. 2010.

52 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 5, May 2011


Siamak Janianpour received the B.Sc. degree in
mechanical engineering with honors from the
Islamic Azad University Tehran south branch,
Tehran, Iran, in 2008. He is now M.Sc. student in
electrical and electronic engineering at Sabzevar
Tarbiat Moallem University in Iran. His current
research interests are computer vision, pattern
recognition, digital image processing and analysis,
intelligent systems, intelligent healthcare systems
and etc. He is a member of the IEEE.
Hossein Ghayoumi Zadeh received the B.Sc.
degree in electrical engineering with honors from
Sabzevar Tarbiat Moallem University, Sabzevar,
Iran, in 2008. He is now M.Sc. student in electrical
and electronic engineering at Sabzevar Tarbiat
Moallem University in Iran. His current research
interests include computer vision, pattern
recognition, image processing, artificial neural
network, intelligent systems, fuzzy logic and soft
computing and etc.
Javad Haddadnia received his B.Sc. and M.Sc.
degrees in electrical and electronic engineering
with the first rank from Amirkabir University of
Technology, Tehran, Iran, in 1993 and 1995,
respectively. He received his Ph.D. degree in
electrical engineering from Amirkabir University
of Technology, Tehran, Iran in 2002. He joined
Tarbiat Moallem University of Sabzevar in Iran
since 2002 as an associated professor. His research
interests include neural network, digital image
processing, computer vision and medical
Engineering. He has published several papers in
these areas. He has served as a Visiting Research
Scholar at the University of Windsor, Canada
during 2001- 2002. He is a member of SPIE,
53 http://sites.google.com/site/ijcsis/
ISSN 1947-5500

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master Your Semester with a Special Offer from Scribd & The New York Times

Cancel anytime.