G010413945

IOSR Journal of Electronics and Communication Engineering (IOSR-JECE)
e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 10, Issue 4, Ver. I (Jul - Aug .2015), PP 39-45
www.iosrjournals.org
DOI: 10.9790/2834-10413945 www.iosrjournals.org 39 | Page
SDSELMS-A Simplified Adaptive Channel Estimation Algorithm
for MIMO-OFDM System
Madhuri P. L., Anu Anna John
(Department of Electronics and Communication Engineering,
Mar Baselios College of Engineering and Technology, Trivandrum, India)
Abstract : The demand for high speed wireless applications increased the popularity of MIMO-OFDM
systems, which provides high data rates with improved link reliability and capacity. One of the key challenges
faced by a MIMO-OFDM system is accurate Channel Estimation (CE). Adaptive Channel Estimation (ACE) is
widely used for CE as it can track the time varying wireless channel rapidly. Among different ACE algorithms
Least Mean Square (LMS) algorithm is the one with low complexity. But it suffers from high Mean Square
Error. Normalized LMS algorithm overcomes this disadvantage with low MSE but the complexity is very high.
Main design aim of any communication system is to minimize the complexity. In order to further reduce the
complexity of LMS algorithm, several simplified algorithms are used. It includes Sign Error LMS (SELMS)
algorithm, Sign Data Normalized LMS (SDNLMS) algorithm etc. In this paper, a new, low complex Sign Data
Sign Error LMS (SDSELMS) algorithm is proposed which improves the convergence rate and reduces the
computational complexity of the existing algorithms which is required for making channel estimation simpler.
Keywords - Channel estimation, MIMO-OFDM, LMS, NLMS, SDNLMS
I. Introduction
Wireless Communication Technology has developed many folds over the past few years. This
developments lead to the increasing demand for high speed wireless communication systems. In order to meet
this increasing demand for various multimedia and internet related services; an efficient, high speed wireless
communication technology is required. The Multiple-Input Multiple-Output (MIMO) system combined with the
multicarrier modulation scheme, Orthogonal Frequency Division Multiplexing (OFDM) is such a popular and
effective communication system that provides large system capacity and high data rates without any extra
consumption of bandwidth and power [1].
Multiple-input multiple-output system refers to the use of multiple antennas at both the transmitter and
receiver ends. It increases the system capacity by exploiting the spatial dimension. A MIMO system transmits the
signal through different antennas without using any extra bandwidth or power than that required for a single-input
single-output (SISO) system. It provides increased link reliability, quality of service, improved coverage and
performance by mitigating the effects of fading.
Orthogonal frequency division multiplexing is a multi-carrier modulation technique, which ensures high
data rate transmission in delay-dispersive environment by eliminating inter symbol interference. An OFDM
system converts a frequency selective channel into several parallel frequency flat channels and thus reduces the
effect of fading and makes multi channel equalization simpler. At first it converts the high speed serial data
stream into a number of low speed data strings that can be transmitted in N subchannels. Then the OFDM
technology modulates them using N subcarriers that are orthogonal to each other. The N modulated signals are
transmitted together to the destination [2].
MIMO in combination with OFDM is widely used due to its best performance in terms of high data
rates, capacity and good outcome in frequency selective fading channels. When the generated OFDM signal is
transmitted over a number of antennas in order to achieve higher transmission rates or to achieve diversity then
the system is referred to as a MIMO-OFDM system. Together they overcome the effects of ISI, improve the link
reliability and also improve the spectral efficiency by using orthogonal subcarriers. So MIMO OFDM is a
promising candidate technology for modern high speed wireless communication [3].
Channel estimation is one of the major challenges faced by a MIMO-OFDM system. It refers to the
process of characterizing the effect of physical medium on the input sequence. Channel estimates or Channel
State Information (CSI) are required at the receiver for the effective retrieval of transmitted signal. Various
channel estimation techniques are used to judge the physical effects of the medium present for MIMO-OFDM
systems [4-6]. As the wireless channel is rapidly varying with time, adaptive channel estimation (ACE)
techniques that are able to update the parameters continuously, are suitable for the estimation of such a time
variant channel. They are pilot based channel estimation techniques, which estimate the channel by transmitting a
sequence of known signals along with the data signals. Main advantage of ACE techniques is that it doesn’t
require any prior knowledge of channel and noise statistics [7]. Only parameter required is the knowledge of the

SDSELMS-A Simplified Adaptive Channel Estimation Algorithm for MIMO-OFDM System
received signal, which can update the channel parameters automatically. Out of various adaptive channel
estimation algorithms, Least Mean Square (LMS) algorithm is the simplest one with low complexity [8]. The
three important parameters that are to be considered in selecting the adaptive algorithms for channel estimation
are Mean Square Error (MSE), computational complexity and convergence rate [9]. Thus the main aim of any
channel estimation technique is to minimize the MSE by utilizing as little computational resources as possible
allowing easier implementation with low complexity and faster convergence.
In order to further reduce the complexity of LMS algorithm, several simplified LMS algorithms can be
used. It includes the Sign Error LMS (SELMS) algorithm, Sign Data Normalized LMS (SDNLMS) algorithm
[10] etc. In this paper a low complexity Sign-Data Sign-Error Least Mean Square (SDSELMS) algorithm is
proposed for channel estimation of MIMO OFDM System, which further reduces the complexity of other
simplified algorithms.
II. MIMO-OFDM System Model
The MIMO OFDM system model based on pilot based channel estimation is depicted in Fig. 1.
Figure 1. MIMO-OFDM system model
The input data in the form of binary sequence is first mapped into multi amplitude multi phase signals
according to the type of modulation used. Here QPSK modulation is used. The modulated serial data sequence is
then converted into N parallel data sequences with the help of a serial to parallel converter. Then pilot signals are
inserted uniformly between the information sequences at known positions for the purpose of channel estimation at
the receiver side. After inserting the pilot signals, the parallel data sequences are passed through an IFFT block
which converts the complex data sequence in frequency domain into time domain. In the next stage cyclic prefix
is added to each of the OFDM signal. Cyclic prefix is the copy of last part of OFDM symbol that is added at the
beginning of each OFDM symbol. It is inserted mainly to prevent the effect of Inter Symbol Interference (ISI) in
OFDM signals. It acts as a buffer between two OFDM symbols and thus prevents ISI. After adding cyclic prefix,
the parallel sequence is converted into serial data sequences and is transmitted through a number of antennas to
the receiver through a Rayleigh fading channe [11l.
At the receiver end the data is received by a number of receive antennas as it is a MIMO system. The
received serial data is converted into parallel sequences with the help of a serial to parallel converter. The cyclic
prefix added at the transmitter is removed and the received signal is passed through an FFT block for
demultiplexing the multicarrier signals. It converts the time domain signal back into frequency domain signal.
The pilot signals are then extracted from demultiplexed sample. This pilot signal is further used for the estimation
of channel using various adaptive channel estimation algorithms. After estimating the channel, transmitted data
signals can be recovered by dividing the received signal by the obtained channel frequency response. Further the
parallel data is converted into serial data sequence and is demodulated to reconstruct the original binary data at
the receiver output [12].
III. Adaptive Channel Estimation Algorithms
As the wireless channel is rapidly varying with time, a channel estimation algorithm that can
automatically adapt according to time is required for effective channel estimation. An adaptive algorithm is such
a process that can update its parameters automatically as it gain information about the changing environment [11].
An Adaptive estimator is able to update the parameters without any prior knowledge about the channel and noise
statistics. The only requirement is the knowledge of the received signal.

Adaptive CE algorithms use an adaptive filter to estimate the MIMO OFDM channel. Fig. 2 shows an
adaptive channel estimation scheme.
Figure 2. Adaptive Channel Estimation Scheme
In Fig. 2, )(ix represents the transmitted random sequence and  )(iy represents the random sequence
received at any time instant i after passing through the channel. The sequence  )(ix is passed through a
Rayleigh channel which is represented by the unknown time variant system with channel impulse response
sequence h . The output sequence at the receiver can be described by:
)()( inhxiy i  (1)
where, )(in represents the Additive White Gaussian Noise (AWGN) with mean zero. Thus the received
signal )(iy includes the effect of channel h and the AWGN noise )(in .
At the receiver side consider an FIR filter with adjustable weights )(iw which is excited by the same
pilot sequence used at the transmitter. This FIR filter can mimic the unknown channel by updating the weight
vectors. FIR filter output can be represented as:
i
w
i
xiy )(ˆ (2)
where, iw denotes the filter weight vector at i th
instant of time. The error between the desired output
and filter output is given by:
ii wxiyie  )()( (3)
The FIR filter coefficients are adjusted from iw to 1iw with the help of this error signal. The error
signal is minimized through a number of iterations. After a number of iterations the output of the adaptive filter
)(ˆ iy will assume values close to the output through the actual channel )(iy . Thus the adaptive filter will behave
similar to the unknown channel [10]. Therefore, we can recover the channel impulse response sequence, h by
determining the optimum value of weight vectors 1iw .
There are a number of adaptive CE algorithms available with varying performance. Among them, the
algorithm which is most popular and with lowest complexity is the Least Mean Square (LMS) algorithm.
3.1 LMS ACE Algorithm
The LMS algorithm is an adaptive algorithm which incorporates an iterative procedure that makes
successive corrections in the weight vector in the direction of the negative gradient vector which eventually leads
to the minimum mean square error (MSE) compared to other algorithms [11]. LMS algorithm is relatively simple
with the lowest complexity. In LMS algorithm the weight vector is updated by using the equation:
 iiiii wxiyxww  )(*
1  (4)
where, µ is the step size parameter.
x(i)
n(i)
y(i)
)(iy

Unknown
time variant
system h
FIR filter
model wi
Adaptive
Algorithm
e(i)

3.2 NLMS ACE Algorithm
The main drawback of the standard LMS algorithm is its sensitivity to the scaling of inputs. This makes
it hard to select a step size µ that guarantees the stability of the algorithm. Normalized LMS (NLMS) algorithm is
a variant of LMS algorithm that solves this problem by normalizing the update factor with the power of the input
signal [12]. It is similar to the LMS algorithm except the weight update calculation which is given by:
 iii
i
ii wxiyx
x
ww 

 )(*
21


(5)
where,  is a small positive constant to avoid division by zero.
IV. Simplified ACE Algorithms
Some of the adaptive filter applications require us to implement the adaptive filter algorithms on
hardware targets like DSP devices, application specified integrated circuits and FPGA targets. A simplified
version of standard LMS algorithm is required for such targets. The LMS algorithms can be simplified with the
help of signum function. Signum function is defined as,
+1 if x > 0
sign(x) = 0 if x = 0 (6)
-1 if x < 0
Applying sign function to the standard LMS algorithm returns the following simplified sign LMS
algorithms:
4.1 SELMS ACE Algorithm
Sign Error Least Mean Square (SELMS) algorithm applies the signum function to the error signal. When
the error signal is zero, this algorithm does not involve any multiplication operation. When the error signal is non
zero, this algorithm involves only one multiplication operation. SELMS algorithm updates the weight vector of
the adaptive filter using the equation,
 iiiii wxiysignxww  )(*
1  (7)
4.2 SDNLMS ACE Algorithm
Sign Data Normalized Least Mean Square (SDNLMS) algorithm applies the signum function to the
input data signal [10]. When the input signal is zero, this algorithm does not involve any multiplication. When the
input signal is non zero, this algorithm involves only one multiplication operation. SDLMS algorithm updates the
weight vector of the adaptive filter using the equation:
   iii
i
ii wxiyxsign
x
ww 

 )(*
21


(8)
4.3 Proposed Algorithm: SDSELMS Algorithm
Sign Data Sign Error Least Mean Square (SDSELMS) algorithm applies signum function to both the
input data signal and the error signal. It is a combination of SDLMS and SELMS algorithms. If either the input
signal or the error signal is zero, this algorithm does not involve any multiplication. The algorithm involves a
multiplication only if both the input signal and the error signal is non zero. Thus it reduces the number complex
multiplications required. SDSELMS algorithm updates the weight vector of the adaptive filter using the equation:
   iiiii wxiysignxsignww  )(*
1  (9)
Thus by applying signum function to both the data signal and the error signal, the SDSELMS algorithm
further simplifies the complexity of the low complex SDNLMS algorithm.
V. Experimental Results
The main performance parameters for a channel estimation algorithm are Mean Square Error (MSE),
Convergence rate and Computational complexity. This section compares the MSE performance and
computational complexity for LMS, NLMS, SELMS, SDNLMS, SDSELMS algorithms for MIMO OFDM
systems. Simulation is done in MATLAB 2013. A 2*2 MIMO-OFDM system is considered in these simulations.
Modulation scheme used is QPSK. Noise used in the simulation is AWGN noise and the channel under
consideration is Rayleigh fading channel.

5.1 MSE performance
Fig. 3 shows the MSE performance for the various ACE algorithms described in the above section.
Among the algorithms, NLMS algorithm provides the best MSE performance. SDNLMS algorithm provides
almost same MSE performance as that of NLMS algorithm. MSE performance of SDSELMS algorithm is less
than SDNLMS. SELMS algorithm has the least MSE performance. LMS algorithm also has less MSE
performance.
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-140
-120
-100
-80
-60
-40
-20
0
20
Iterations
MeanSquareErrorindB
LMS
NLMS
SELMS
SDNLMS
SDSELMS
Figure 3. MSE Performance of different ACE algorithms
5.2 Convergence Rate
From the simulation results it is seen that Sign Data Sign Error LMS CE algorithm provides the fastest
convergence among all the CE algorithms. SELMS CE also provides faster convergence. The LMS algorithm
comes next. NLMS algorithm has the least convergence rate. SDNLMS algorithm has better convergence when
compared with NLMS algorithm. Thus SDSELMS algorithm provides good performance in terms of
convergence.
The convergence performance of the new SDSELMS algorithm is shown in Fig. 4. Fig. 5 shows the
comparison of convergence performance for the various existing ACE algorithms with the new SDSELMS
algorithm.
0 20 40 60 80 100 120 140 160 180 200
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Iterations
MeanSquareError
Proposed Algorithm : SDSELMS
Figure 4. Convergence Performance of new SDSELMS algorithm

0 50 100 150 200 250 300 350 400 450 500
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Iterations
MeanSquareError,MSE
LMS
NLMS
SELMS
SDNLMS
SDSELMS
Figure 5. Comparison of Convergence Performance for different ACE algorithms
5.3 Computational Complexity
Computational complexity is one of the key parameter that is to be considered while designing a channel
estimation algorithm. The complexity of the CE algorithm should be as low as possible for reducing the
complexity of the whole system. The computational complexity in terms of the number of complex
multiplications, divisions, additions and sign operations required during each iteration for an L length complex-
valued data, is shown in Table I.
Table I
Computational complexity for different ACE algorithms
Method Addition Multiplication Division Sign
LMS 2L 2L+1 - -
SELMS 2L 2L - 1
NLMS 3L 3L+1 1 -
SDNLMS 3L 2L+1 1 1
SDSELMS 2L 2L - 2
From the calculations it can be seen that SDSELMS algorithm has the least complexity as it requires less
number of complex multiplications. It is due to the use of two signum functions. SELMS and LMS algorithms
also have low complexity but greater than SDSELMS algorithm. NLMS algorithm has the highest complexity as
it involves more number of complex multiplications and a division operation. SDNLMS has comparatively less
complexity than NLMS algorithm as it requires less number of complex multiplications than NLMS algorithm.
VI. Conclusion
MIMO-OFDM technology has became a vital part in modern wireless communication systems which
provides significant increase in data rates, capacity and link reliability without any additional bandwidth or power
consumption. Channel estimation is one of the key challenges faced by MIMO OFDM systems. In this paper a
new SDSELMS ACE algorithm is proposed for the channel estimation in MIMO-OFDM Systems. Its
performance is compared with some low complexity adaptive channel estimation algorithms like LMS, NLMS,
SELMS, SDNLMS. From the experimental results and analysis, it can be seen that the proposed SDSELMS
algorithm provides the lowest complexity and faster convergence than the other ACE algorithms. So for those
applications which require low complexity and faster convergence SDSELMS algorithm is an effective choice.

References
[1] Vaishali B. Niranjane, Nivedita A. Pande, And Anjali A. Keluskar, “Adaptive Channel Estimation Technique For MIMO-OFDM,”
International Journal Of Emerging Technology And Advanced Engineering, ISSN 2250-2459, Vol 3, Issue 1, January 2013.
[2] S. A. Hosseini, And S. A. Hadei, “Low-Complexity Adaptive Channel Estimation For MIMO-OFDM Systems Over Rayleigh
Fadingchannels,” National Conference On New Idea On Electrical Engineering, Vol. 12, March 2012.
[3] Pei-Sheng Pan And Bao-Yu Zheng, “An Adaptive Channel Estimation Technique In MIMO -OFDM Systems,” Journal Of
Electronic Science And Technology Of China, Vol. 6, No. 3, September 2008.
[4] M. M. Rana, And M. K. Hosain, “Adaptive Channel Estimation Techniques For MIMO OFDM Systems,” International Journal Of
Advanced Computer Science And Applications, United States, Vol. 1 No. 6, December 2010.
[5] D. B. Bhoyar, Dr. C. G. Dethe, Dr. M. M. Mushrif, A. P. Narkhede, “Leaky Least Mean Square (LLMS) Algorithm For Channel
Estimation In BPSK-QPSK-PSK MIMO-OFDM System,” International Conference On Automation, Computing, Communication,
Control And Compressed Sensing (Imac4s), Pp.623 – 629, March 2013.
[6] T. Chang, W. Chiang, And Y. P. Hong, “Training Sequence Design For Discriminatory Channel Estimation In Wireless MIMO
Systems,” IEEE Transactions On Signal Processing, Vol. 58, No. 12, Pp. 6223–6237,2010.
[7] S. Haykin, “Adaptive Filter Theory,” Prentice-Hall International Inc, 1996.
[8] A. H. Sayed, “Fundamentals Of Adaptive Filtering,” John Wiley & Sons, June 13, 2003.
[9] Sinem Coleri, Mustafa Ergen, Anuj Puri And Ahmad Bahai, “Channel Estimation Techniques Based On Pilot Arrangement In
OFDM Systems,” IEEE Transactions On Broadcasting, Vol. 48, No. 3, September 2002.
[10] T. A. Dewan, S. Hasan And F. Hossain, “Low Complexity SDNLMS Adaptive Channel Estimation For MIMO-OFDM Systems,”
International Conference On Electrical Information And Communication Technology (EICT), Pp.1-5, 2013.
[11] A. Zaier And R. Bouallègue, “Channel Estimation Study For Block - Pilot Insertion In OFDM Systems Under Slowly Time
Varying Conditions,” International Journal Of Computer Networks & Communications (IJCNC), Vol.3, No.6, November 2011.
[12] N. Daryasafar, A. Lashkari, B. Ehyaee, “Channel Estimation In MIMO-OFDM Systems Based On Comparative Methods By LMS
Algorithm,” International Journal Of Computer Science Issues (IJCSI), Vol. 9, Issue 3, No 3, May 2012.

G010413945

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to G010413945 (20)

More from IOSR Journals (20)

Recently uploaded (20)

G010413945