Comparative Analysis between Direction of Arrival Algorithms

Muhammad Umair Shahid1, Muhammad Nauman2, Danyal Haider3, Yasal Imran4

1
Electrical Engineering Department, Bahria University, Islamabad, Pakistan
1
[email protected], [email protected]
3
[email protected], [email protected]

Abstract: In today¶s world Smart antenna systems have pattern of the antenna is changed by altering the weights of
emerged as one of the most prominent way for an efficient antenna array by a beam forming algorithm.
wireless system. According to authors, in this research a
comparative study has been conducted among 3 direction of
arrival algorithms namely Capon, MUSIC and Pisarenko.
After application of Matlab simulations on the mentioned
algorithms using user capacity and interference as
benchmark parameters compared by their pseudo spectrum,
Pisarenko estimation has been found to cover most of the
users with comparatively low interference than the former
two.

Keywords: Smart Antenna; Direction of Arrival (DOA);

Capon; MUSIC; Pisarenko', Sensor arra;, Beam forming;
user capacity; interference.

I. INTRODUCTION
Fig. 1. Smart antenna system (Block Diagram)
As in the current scenario, there is a huge demand on
Communication Vendors and Mobile Operators to give high ,Q)LJXUH6NGHQRWHVWKH QXPEHURI,QFRPLQJ6LJQDOVࣝ
speed voice and data services to their users. But at the same denotes the direction of incoming signals. X(k) denotes the
time, these operators support more users on same base station Array Element and W is denoting the Antenna Weights.
due to huge increase in the users. This helps the operators to
make the overall network very cost effective. As a A. CATEGORIES OF SMART ANTENNA
consequence user suffers a lot, because the overall broadcast
spectrum becomes limited. As the operators attempt to There are three main categories of Smart Antenna:
enhance the network traffic within the fixed bandwidth, it
1. SIMO (Single Input Multiple Output): transmitter end
creates Interference issues and degrades the quality of signal
uses only one antenna but two or more antennas are used
due to which one can suffer from signal distortion on the
at receiver end.
receiving end [1] [2].
2. MISO (Multiple Input Single Output): at transmitter end
One solution to cater this problem is to use spatial diversity
multiple antennas are used but only one antenna is used at
techniques. Smart antenna systems are one of the most
receiving end.
efficient method of spatial diversity technique. Smart antenna
systems adapt to the changing environment of incoming signal 3. MIMO (Multiple Input Multiple Output): at both
by tracking them down and directing the antenna radiations to transmitter and receiver end, multiple antennas are used.
the desired location (Smart Antennas with MATLAB). The
technology of smart antenna is based on antenna array B. ESTIMATION TECHNIQUES OF SMART ANTENNA
changing the radiation pattern of antenna by adjusting relative
phase and amplitude of antenna array elements [3]. Figure 1 There are two major Estimation Techniques in smart antenna
refers to a smart antenna system, S(K) shows the incoming system:
PRQRFKURPDWLF VLJQDOV ZLWK इ' GLUHFWLRQ WKDW DUH LQFLGHQW
on the antenna array while x(k) denotes antenna array 1. DOA (Direction of Arrival) Estimation
elements. These signals are then multiplied by the incident
array element weight and fed to the DSP processor for Smart antenna system estimates direction of incoming signal
Direction of Arrival (DOA) Estimation. The direction of by using different algorithm such as Music, Pisarenko, Capon
incoming signal is estimated over here and the radiation and Bartlett DOA estimation [4]. The output from DOA

‹,(((

 2017 International Conference on Infocom Technologies and Unmanned Systems (ICTUS'2017),
Dec. 18-20, 2017, ADET, Amity University Dubai, UAE

estimation algorithm is used as a seed to create a directive estimation algorithms. In this paper research is focused on
beam toward the desired location. It also reduces the three main DOA estimation algorithms as under.
Interference and increase the radiated power towards the user.
DOA estimation algorithms takes various parameters into A. MUSIC
account such as array element spacing and number, SNR,
number of desired users and amount of sampled signals. MUSIC is the Abbreviation of MUltiple SIgnal Classification.
Direction of Arrival (DOA) estimation are categorized into Music is the most popular DOA algorithm in finding the High
Resolution. The most primary aspect in MUSIC algorithm is
two separate methods.
Covariance Matrix because this algorithm breaks the
x Conventional Method: Also known as the classical Covariance Matrix into two different orthogonal matrix, such
method which contains further two algorithms, Capon and that noise subspace and the signal subspace. In this algorithm,
the Bartlett. Main drawback in conventional methods is there is an assumption that noise is highly uncorrelated in
that they don¶t have the capability to show the largest every channel and this makes the covariance Matrix Diagonal
number of Angular Resolution of their users [5]. That¶s [7] [8] [9].
why, there is a need of another method which can shows ૚
the highest number of Angular Resolution of their users ࡼࡹࢁ ሺࢋ࢐᫜ ሻ ൌ ૛ (1)
and due to this reason we come to Subspace Methods [6] σࡹ ࡴ
࢏స࢖శ૚หࢋ ࢜࢏ ห
[4].
B. Pisarenko's
x Subspace Method: These algorithms are better in
performance as compared to conventional algorithms. This method was first discovered in 1973 by Vladilen
This method depends on the decomposition of Covariance Fedorovich Pisarenko while he was doing his research on how
Eigen Matrix into the Noise Subspace and the Signal to estimate frequencies of Complex Signal which is present in
Subspace. This method is used to find the Spatial the white noise. This Pisarenko Harmonic Decomposition
Spectrum of Antenna, and further then the Spatial Method reduces the (MSE) Mean Square Error of Output
Spectrum Peaks are used to find Direction of Arrival [4]. Array with the condition that the Norm Vector should equal to
Unity. Eigen Vector that reduces the Mean Square Error
2. Beam forming. correspond to the smallest Eigen value of the cross Correlation
After the direction of arrival estimation of the incident signal, of Output Array [5] [7].
a DSP processor then changes the weights of the antenna to ૚
direct the beam of antenna to the desired incident signals. ࡼࡼࡴࡰ ሺˁሻ ൌ ૛ (2)
ഥࡴ ሺˁሻࢋ
หࢇ തതതห
૚
II. DOA ESTIMATION ALGORITHMS
C. Capon
As the first part of the Smart Antenna System is DOA
estimation, this paper is focused on these algorithms. Many Capon estimation is improved version of Bartlett Algorithm.
researchers have tried to find the best DOA algorithm using The drawback in Bartlett Algorithm was that it fails in
their own defined scenarios and benchmarks. The authors of separating the sources which are located close to the
[13] have analyzed a few of these algorithms namely MUSIC beamwidth [10]. After this, Capon comes which gives the
and ESPRIT varying array size and noise variance. They method of Maximum Likelihood to solve minimum variance
concluded their results by nominating MUSIC algorithm the [8].
most efficient. In [11], the author focused on the application of
૚
MUSIC, ESPRIT and ROOT MUSIC. The author observed ࡼ࡯ ሺˁሻ ൌ ష૚ (3)
the stated algorithms under SNR variance and found out that ࢇ തതതതത
ഥࡴ ሺˁሻࡾ ࢞࢞ ഥሺˁሻ
ࢇ
MUSIC has the best resolution.
III. COMPARISON TABLE BETWEEN DOA
Another paper [14] addresses the similar issue with more METHODS
algorithms then [13] with the same benchmarks. The authors Table 1 shows a comprehensive analysis between different
of this paper have done a comprehensive analysis on all of the DOA algorithms.
algorithms and established that Linear Prediction Method
produces better results than other algorithms under TABLE 1: Comparison between DOA Algorithms
consideration with defined benchmarks.
Algorithms Advantages Disadvantages
All of these papers focused on the noise variance and array Bartlett Algorithm
elements. In this paper we took another approach and The Implementation
fails in separating
introduced the number of user as a benchmark to study DOA Bartlett of Bartlett is very
the sources which
simple and Easy [5].
are located close to



 2017 International Conference on Infocom Technologies and Unmanned Systems (ICTUS'2017),
Dec. 18-20, 2017, ADET, Amity University Dubai, UAE

Algorithms Advantages Disadvantages parameters were kept constant throughout the simulations,
the beam width. which are
This Algorithm is
x Array elements - 12
dependent on the
Array size. x Noise variance ± 0.1
Less Efficient as
compared to Capon x Element spacing ± Ȝ
[5]. Capon DOA estimation for multiple users
5
Capon reduces the array elements=12 Users=4
drawback of Bartlett Capon Algorithm array elements=12 Users=8
array elements=12 Users=12
0
and it gives the was too much
method of Maximum dependent on
Likelihood to solve Sources, That¶s why -5

|P(T)| (dB)
minimum variance. the high correlation
Capon
The Resolution of of sources gives -10
Capon is better than poor results.
that Bartlett. Its performance is -15
More Efficient as limited by Sensor
compared to Capon Noise Power [5].
-20
[5].
-30 -20 -10 0 10 20 30
This Method reduces Angle
In this method, Peaks the Mean Square
represent the roots of Error (MSE) of the Fig. 2. Capon Estimation Results
Pisarenko Polynomial which is Output Array with
in the denominator the condition that the Figure (2) shows simulation results for Capon estimation
[5]. Norm Vector should using the above stated parameters. As we increased the
equal to Unity [5]. number of users the pseudo spectrum for Capon estimation
Its Performance is becomes flat. Capon estimation cannot track more users with
lower as compared accuracy. When the users are increased and their distance
MUSIC is the most to ESPRIT. from each other is not as significant as their number this
popular DOA It is very sensitive to estimation method fails to efficiently track them individually.
MUSIC algorithm in finding the Coherent
Pisaranko DOA estimation for multiple users
the High Resolution Multipath.
0
[5]. It is also very array elements=12 Users=4
array elements=12 Users=8
sensitive to Phase array elements=12 Users=12
-50
and Gain Errors [5].
Signal to Noise -100
Ratio is
Root Music
|P(T)| (dB)

comparatively low -150

Algorithm can only be
Root [11].
applied on Uniform
MUSIC Root finding is less -200
Linear Array (ULA)
accurate because it
[5] [11].
may suffers from -250

loss [5].
-300
It require multiple
ESPIRIT has very
images. -30 -20 -10 0 10 20 30
High Resolution [12]. Angle
ESPRIT It performance is
Its Array calibration is
limited by array
Non Critical [5]. Fig. 3. Pisaranko Estimation Results
geometry [5] [12].
Pisaranko DOA estimation provided best results under the fore
IV. SIMULATION AND RESULTS mentioned conditions. When users were increased from 4 to
This paper focuses on three of the stated DOA estimations 12, the pseudo spectrum almost covered all the users
algorithms. The simulation tool used was Matlab. Some of the accurately. Both of the previous methods failed to persevere



 2017 International Conference on Infocom Technologies and Unmanned Systems (ICTUS'2017),
Dec. 18-20, 2017, ADET, Amity University Dubai, UAE

their accuracy with high number of users unlike Pisaranko [2] M. W. Amrita Soni, "Comparitive Analysis of DOA and
estimation method. Beamforming Algorithms for Smart Antenna Systems",
International Journal of Research in Engineering and
Music DOA estimation for multiple users Technology, vol. 04, no. 05, pp. 468-473, 2015.
5 [3] V. K. K. &. M. S. T. B. Lavate, ³Performance Analysis of
array elements=12 Users=4 Music and Esprit DOA Estimation Algorithms for Adaptive
array elements=12 Users=8
Array Smart Antenna in Mobile Communication", in
array elements=12 Users=12
0 International Conference on Computer and Network
Technology, IEEE Computer Society, 2010, pp. 309-311.
[4] S. S. Y. M. Ei Thinzar Aung, ³Performance Comparison of
-5 DOA Estimation Algorithms for Smart Antenna", International
|P(T)| (dB)

Journal of Electronics and Computer Science Engineering, vol.

3, no. 2, pp. 167-174, 2012.
-10 [5] S. K. Singh Neelima, "A Review of Direction of Arrival (DOA)
Estimation for Smart Antenna Structure", Journal Of
Information, Knowledge and Research in Electronics and
-15 Communication, vol. 3, no. 2, pp. 1141-1146, 2014.
[6] A. J. J. Gaurav Chaitanya, "Performance Analysis of Direction
of Arrival Estimation Algorithms for Smart Antenna for
-20 Mobile Communication Systems", International Journal of
-30 -20 -10 0 10 20 30
Angle Scientific & Engineering Research, vol. 3, no. 7, pp. 1-5, 2012.
[7] F. B.Gross, "Angle of Arrival Estimation", in Smart Antenna
with MATLAB, Mc Graw Hill Education, 2015, pp. 7.1.1 -
Fig. 4. MUSIC Estimation Results
7.3.9.
Figure 4 depicts Music DOA estimation results for stated [8] S. S. M. F. Youssef Khmou, "Comparative Study between
parameters. MUSIC estimation fails to keep up with the Several Direction of Arrival Estimation Methods", Journal of
Telecommunications and Information Technology, vol. 1, no.
increase in users. This estimation gave better results with 4
1, pp. 41-48, 2014.
users than 8 and 12 users. This estimation scheme is widely
[9] A. J. Gaurav Chaitanya, " Performance Analysis of Direction of
used but it fails to deliver good results when users are more Arrival Estimation Algorithms for Smart Antenna for Mobile
than 2. Communication Systems", International Journal of Scientific &
Engineering Research, vol. 3, no. 7, 2012.
V. CONCLUSION [10] K.-H. G. Y. Narita, " High-resolution wave number spectrum
This paper deliberates and differentiates among pseudo using multi-point measurements in space ± the Multi-point
Signal Resonator (MSR) technique", Annales Geophysicae,
spectrums of three Direction of Arrival estimation methods for
vol. 29, pp. 351-360, 2011.
Smart Antenna Systems keeping user capacity as benchmark.
[11] T. S. Dhope, "Application of MUSIC, ESPRIT and ROOT
Simulation results showed that Capon estimation has MUSIC in DOA Estimation", World Journal of Science and
marginally enhanced results than MUSIC estimation in terms Technology, vol. 1, no. 8, pp. 20-25, 2011.
of user capacity. MUSIC estimation cannot locate more than [12] R. B. Deven Pradhan, "Direction of Arrival Estimation via
two users with less distance. Pisaranko estimation technique ESPRIT Algorithm for Smart Antenna System", International
produces best results than Capon, which is better than MUSIC Journal of Computer Applications, vol. 118, p. 6, 2015.
estimation technique, even when number of users increase. [13] Umar Mujahid, ³Spectral Estimation for Smart Antenna
Systems´, International Conference on Computers, Controls
REFERENCES and Communication (IC4), 2013.
[14] Umar Mujahid, ³Performance Analysis of Spectral Estimation
[1] A. N. Jadhav, V. M. Mhalgi, S. D. Bhosale, P.B. Khapale,
for Smart Antenna System´ World Applied Science Journal,
³Evaluation of PM and MUSIC: Direction of Arrival
vol 27, issue 12, pp. 1695-1705, 2013.
Estimation for Smart Antenna System", International Journal of
Engineering, Economics and Management, vol. 1, no. 1, pp. 1-
4, 2012.