Hybrid Adaptive Channel Estimation Technique in Time and Frequency Domain for MIMO OFDM Systems

International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7, Issue-5C, February 2019
105
Published By:
Blue Eyes Intelligence Engineering
& Sciences PublicationRetrieval Number: E10260275C19/19©BEIESP
Abstract--- A multiple-input multiple (MIMO) communication
model is integrated to orthogonal frequency division multiplexing
(OFDM) for reliable and data transmission at a higher rate in
the broadband wireless channels. A significant part in the
wireless data transmission is use of adaptive channel estimation
techniques in which the channel is frequently varying. This paper
introduces a hybrid adaptive channel estimation technique by
integrating the beneficial characteristics of the time domain as
well as frequency domain. The column based time domain and
the row based frequency domain are combined together and they
are used based on the channel quality and received bit error rate
(BER). Here, the channel estimation at pilot frequencies depends
on minimum mean square error (MMSE). For the evaluation of
the proposed techniques, a set of experiments and detailed
comparative analysis is made interms of different measures
under different condition. The experimental results depicted the
proposed method outperforms the other methods.
Keywords--- Channel estimation, OFDM, CSI, MMSE.
I. INTRODUCTION
Adaptive resource allocation using the channel state
information (CSI) received more attention of improving data
rate by achieving high quality of service in time-varying
wireless channels [1]–[4]. So, the adaptive modulation is
introduced in which the rate of data transmission and power
are altered using perfect CSI at the transmitter which
indicates the results are enhanced when compared to static
wireless communication through flat fading channels [5]–
[6]. In the meantime, frequency selection in the wireless
channel will be inevitablein case of an increase in
transmission symbol rate. OFDM is an efficient method
which uses the concept by transforming a frequency
selective fading channel to a collection of closely flat
narrowband orthogonal fading channels; hence the
simplified equalization is attained at the receiver [7]. The
main benefits of adaptive modulation techniques used on
precise CSI were assumed for multiple carrier system [3],
[8]. Several works on adaptive modulation indicated that
there is an assumption of proper CSI present at the
transmitter side. However, it is impossible for the perfect
CSI at the transmitter because of time variations in the
channel and improper channel estimation due to noise. The
impact of improper channel estimation on the results of
adaptive modulation using precise CSI at the transmitter is
investigated and depicted high sensitivity to bit error
rate(BER) for channel estimation error as well the variations
Revised Version Manuscript Received on 22 February, 2019
S.B. Lenin, Research Scholar, Sathyabama University, Tamil Nadu,
India. (e-mail: lenin.ram@gmail.com)
N. Tamilarasan, Research Scholar, Pondicherry University, Tamil
Nadu, India. (e-mail: neithalarasu@gmail.com)
S. Malarkkan, Principal, Manakular Vinayagar Institute of
Technology, Pondicherry, India. (e-mail: malarkkan_s@yahoo.com)
in the channel. So, adaptive modulation techniques should
be considered for the use of improper CSI at the transmitter
[9]–[14].
Channel sounding by the use of pilot symbol assisted
modulation (PSAM) [15]–[17] in which the pilot symbols
are appended in a particular time interval while transmitting
the sequence rather than inserting a long training sequence
in prior to the transmission of the signal, is an efficient
model for channel estimation and prediction in fast fading
environment. The two kinds of fundamental channel
estimation methods in OFDM are block type and comb type
pilot-based channel estimation. The former type allocates
the pilot signal to specific OFDM block and transmits it
periodically in the time domain. The latter type distributes
the pilot signals uniformly in every OFDM block. The comb
pilot-based channel estimation uses algorithm for estimating
the channel at the pilot frequencies and interpolates the
channel at data frequencies. The channel estimation at pilot
frequencies depends on least square (LS), minimum mean
square error (MMSE), zero forcing (ZF) or least mean
square (LMS) method, while different interpolation
techniques can be incorporated for the channel estimation at
data frequencies [18, 19]. On the other hand, as pilot symbol
does not carry information, it still consumes a significant
amount of energy and the insertion of pilots produces an
overhead in the transmission model. Hence, an effective
model should consider the optimal usage of pilots and
adaptive systems. An adaptive PSAM technique is presented
in [11] to optimize the pilot parameters for single carrier
transmission using imperfect CSI. To overcome the
limitations of the existing techniques, this paper presents a
novel, hybrid adaptive channel estimation technique by
integrating the beneficial characteristics of the time domain
as well as frequency domain. The column based time
domain and the row based frequency domain are combined
together and they are used based on the channel quality. The
channel estimation at pilot frequencies depends on minimum
mean square error (MMSE). For the evaluation of the
proposed techniques, a set of experiments and detailed
comparative analysis is made interms of different measures
under different conditions. The experimental results
depicted the proposed method outperforms the other
methods.
The paper organization is given as follows: Next section
deals with the system model of a wireless MIMO-OFDM
communication system in time domain and frequency
domain.
Hybrid Adaptive Channel Estimation
Technique in Time and Frequency Domain
for MIMO-OFDM Systems
S.B. Lenin, N. Tamilarasan, S. Malarkkan 

International Conference on Advances in Signal Processing, Power, Embedded, Soft Computing, Communication and
Control Systems (ICSPECS-2019) | 11th & 12th January 2019 | GPREC, Kurnool, A.P. India
106
Published By:
Section 3 explains the presented method and the results
are analyzed in Section 4. At the end, section 5 concludes
the paper.
II. SYSTEM MODEL
In this paper, we consider synchronization of perfect
timing and perfect channel estimation together. Fig. 1 shows
the overall system model employed in this study. Initially,
the input from the source will be modulated and it is
converted by the use of S/P. Then, the pilots are inserted
adaptively and then IDFT takes place. Next, the guard bits
are inserted to avoid overlapping of data. Then, they are
converted back to serial data and are transmitted to the
destination. At the receiving side, initially S/P conversion
takes place and then guard bits are removed which are
included at the transmitter side. Then the demodulation, i.e.
DFT takes place. Then, the hybrid channel estimation takes
(frequency and time domain) by the use of demodulated data
and reference pilot signals. Then, the interference will be
cancelled and the signals are detected by the detector. The
detected signals are sent back for effective channel
prediction and channel estimation. Finally, the P/S again
takes place and the signal will be demodulated.
Fig. 1: Overall system model
With receive and transmit antennas, assume an
OFDM depended MIMO communication system which is
wireless. By the below expression, the transmitted time
domain signal can be demonstrated.
∑ (1)
Where is the symbol sent to the antenna and
subcarrier and belongs to the constellation with size
| is the modulation order,
, is the FFT size, is the sample of the
time domain signal transmitted on the antenna. The
cyclic prefix vector can be demonstrated as:
( ) (2)
where , is the guard interval length.
The sample of the received time domain signal at the
antenna might be equated as:
∑ ∑ ( )
(( )) (3)
where
( )
is the
channel tap gain among receive antenna and the
transmit antenna, (( )) indicates a cyclic shift in the
base of and is a sample of AWGN noise with zero
mean and variance .After, in frequency domain, the
channel coefficients can be written as:
∑ ( )
( )
(4)
∑ ∑
( )
( )
∑ (5)
where is average of the
channel tap. Without loss of generality, only the
subcarrier of the MIMO-OFDM receiver is assumed.
Eq.(4) can be reformulated as:
∑ (( )) (6)
Where ( ( ) ) , { } and
* + are together channel matrices,
( ( ) ) , ( ( ) ) ,
is an i.i.d. complex Gaussian noise with zero mean and
variance . Furthermore, the input data of the
subcarrier as a binary vector,
( ) , where ( ) and
* +. The binary data vector is mapped to the symbol
vector .
III. PROPOSED HYBRID ADAPTIVE CHANNEL
ESTIMATION MODEL
A novel hybrid adaptive channel estimation technique is
introduced by integrating the beneficial characteristics of the
time domain as well as frequency domain. The row based
frequency domain involves less complexity at the cost of
slightly degraded performance. But, the column based time
domain offers higher performance but with the cost of high
complexity. So, in this paper, we integrate the benefits of the
time domain and frequency domain for channel estimation.
At the initial stage, by default, the time domain based
channel estimation is employed. In case of better channel
quality and attained low BER for a predefined time period,
then a swapping of frequency domain from time domain
takes place. Similarly, when the channel quality becomes
worse and the BER is increased, an inverse swapping of
frequency domain to time domain takes place. This adaptive
swapping nature of both the time and frequency will helps to
adaptively approximate the channel quality in an effective
way. The proposed hybrid model investigated the feature of
fading channels and the estimates the channel in time as
well as frequency domain. Here, it is estimated that the CIR
( )as a linear function of the time variable ,
( ) ( ) ( ) (7)
where ( )and ( )are the parameters to be estimated.
Under varying channel conditions with a channel length , it
is needed to estimate 2 parameters for every burst and

107
Published By:
needs to be estimated for each burst, and the CIR of one
burst can be computed easily using the parameters. The
sample received at time can be defined as
( ) (8)
where - are the transmitted
symbols, and , ( ) ( ) (
)- ,for m = 0, 1, with ( ) indicates the transpose
function. Using the known training symbols sent at the
starting and ending of every burst, the received samples
using the training symbols can be represented in a matrix
format as given below
(9)
, - 0 1 (10)
Where
, - ( )
(11)
, - ( )
(12)
[ ]
( )
(13)
and T is a diagonal matrix indicated by
* + (14)
The cost function for an LS criterion can be defined as
follows:
( ) ( ) (15)
where , - , - and ( ) is the
Hermitian transpose operator.
MMSE estimator [30] that minimizes the mean-squared
error can be defined by
( ) ( )
( ( ) ( )
) (16)
Where , - is the autocorrelation matrix of
AWGN vector W. The ZF estimator can be represented as
( ) ( )
( ( ) ( )
) (17)
The MMSE reduces the square error in case of channel
noise is and it becomes the ZF solution in case of the
absence of noise. So, the results of the ZF and MMSE are
identical in case of high. In addition, the MMSE provides
better results when the SNR is low. The ̂, reduced can
be attained from the equation and the solution is
̂ [ ] * + (18)
where . Here, an iterative process will be
carried out for the determination of ̂and it eliminates the
determination of the matrix inversion . And, and are
the ith and jth component of the row and column vector
respectively. Using the above Eqs., theith and jth
received sample of is expressed by
( ) (19)
( ) (20)
Where, and is the ith row and
column of the matrix . The estimation of and , i.e. rows
and columns are almost identical.The computed ̂ at index ,
called ̂( )can be determined by the minimization of the
residual sum of squares represented by ( ) ∑ ,
- [ ] and ( ) is the complex conjugate. Hence,
the computed ̂( )is calculated by as
̂( ) ,∑ - ,∑ - (21)
where ( ) ∑ is a non-negative
definite correlation matrix. For ensuring the correlation
matrix ( )is always positive definite, and nonsingular, it is
defined that ( ) where is a small positive constant
and is the identity matrix; therefore, every
element on the main diagonal of correlation matrix ( )is
included to a small positive constant [17]. Here, by the use
of matrix-inversion lemma, the inverse of matrix ( )can be
determined in an iterative manner
( ) ( ) ( )
( ) ( )
( )
( )
( )
( ) ( ) ( )(23)
Where ( ) (
. ( ) /
. ( ) /
) is the gain
vector. Finally, the computed ̂( )in can be defined as
̂( ) [ ( ) ( ) ( )] ,∑ -
( ) ∑ ( ) ,
( ( ) ∑ )-
̂( ) ( ), ̂( )] ( )
Therefore, it is found that the inversion of the correlation
matrix in (20) got replaced by the inversion of
scalar* ( ) in the estimation of u. While
estimating the parameters and , the CIR of the whole
burst can be easily attained. The above explanation is
mainly focused on time domain and it is identical to
frequency domain. It should be noted that the column
elements are based on time domain and the row elements are
focused on frequency domain.
IV. PERFORMANCE EVALUATION
For the effective investigation on the performance of the
proposed method, an extensive set of experimentation takes
place under varying number of antennas, pilot, SC and
mobility speed. The following parameter settings given in
Table 1 are used for experimentation.
Table 1: Parameter settings
S. No Parameters Value
1. Operating frequency 2.4GHz
2. Modulation QPSK, BPSK
3. Estimation ZF, MMSE, ML
4. No. of symbols 64
5. No. of Sub Carrier 256
6. Channel Model Rayleigh
7. Antennas 2x2

108
Published By:
Fig. 2 provides the comparative analysis of the results
with different number of SC and antennas using the
proposed method. From the Fig., it is obvious that the worse
results are attained for less number of SC and antennas. In
contrast, the results are significantly improved with more
number of SC and antennas. It is verified from the figure
that the BER is high for the 64SC with 2*2 antennas and the
BER is slightly reduced for the 128SC with 2*2 antennas. In
addition, it is noted that the BER is again reduced with an
increase in antennas, i.e., 64 SC with 4*4 antennas
compared to 64SC with 2*2 and 128SC with 2*2. Here, it is
noted that the increase in antenna from 2*2 fails to attain
better results compared to 4*4. Though the number of
subcarriers is low, the increase in receiving antenna from 2
to 4 has great impact on the BER. Similarly, when the
transmitting antennas are also increased from 2 to 4 along
with an increase in subcarriers, i.e. 128 SC with 4*4
antennas, the maximum performance is attained with
minimum BER. These observations indicated that effective
results are attained for an increase in number of SC and
especially number of antennas.
Fig. 2: BER vs.SNR Comparison with 64 and 128
antennas
Fig. 3: Adaptive Channel Estimation for 100km/hr with
2*2 antennas
Fig. 3 shows the detailed comparative analysis of
different channel estimation techniques such as MMSE, ZF
and LS under varying speeds of 100km/hr. This analysis is
used to validate the proposed method interms of high speed
mobility. From this figure, it is clearly observed that the
MMSE is not highly as efficient as ZF. Though ZF attains
better results compared to MMSE under 100kn/hr with 2*2
antennas, the LS shows effective performance over ZF and
MMSE.
4*4 antennas
Fig. 4 depicts the performance of the proposed method
under different channel estimation with a speed of 100km/hr
but with the antennas of 4*4. From the Fig., it is clear that
the ZF obtains higher BER and tends to decrease to a
particular point. The LS shows competitive performance
with LS and leads to better results compared to ZF. And, the
MMSE starts with the significantly lower BER compared to
ZF and LS. It is noted that the MMSE is found to superior
incase of increasing number of antennas from 2*2 to 4*4
under the speed of 100km/hr.
Fig. 5 depicts the results obtained by the proposed method
under different channel estimation with a speed of
200km/hr. From this figure, it is apparent that the MMSE
attains the lowest bit rate and the MMSE shows better
performance than ZF and LS under the high mobility of
200km/hr. In addition, it is noted that the ZF shows worse
performance with a highest BER whereas LS manages to
maintain BER, but not outperform MMSE. It is observed
that MMSE can provide better BER under 200km/hr and
also with low number of antennas 2*2.
2*2 antennas

109
Published By:
Fig. 6 shows the results attained by the proposed method
under different channel estimation with a speed of 200km/hr
but with the increased number of antennas, i.e. 4*4. From
the Fig., it is obvious that the LS method shows poor
performance with a maximum BER and fails to outperform
other methods. At the same time, the LS method depicts
somewhat better performance than ZF with a lower BER
and it starts reducing upto a specific point. The performance
of the LS is gradually better than ZF and also poor than
MMSE. Furthermore, MMSE starts with the better BER
compared to ZF and LS. Interestingly, MMSE maintains to
achieve better results compared to other methods with an
increasing number of antennas from 2*2 to 4*4 under the
speed of 100km/hr.
4*5 antennas
Though the MMSE shows better results under 100km/hr
and 200km/hr, an additional investigation is needed to verify
the consistent results even under high mobility, i.e.
300km/hr. Fig. 7 shows the detailed comparative analysis of
different channel estimation techniques such as MMSE, ZF
and LS under high mobility of 300km/hr. This analysis is
used to validate the proposed method interms of high speed
mobility. From this figure, it is evident that the performance
of the ZF and LS are almost closer to each other and
gradually increases as well as decreases at some point. At
the same time, it is interesting that the MMSE is highly
efficient compared to ZF and LS under 300kn/hr with 2*2
antennas. From these results, it can be seen that the MMSE
is effective in case of high mobility.
2*2 antennas
Fig. 8 provides the performance analysis of the proposed
method under different channel estimation with a high speed
of 300km/hr but with the increased number of antennas from
2*2 to 4*4. From the Fig., it is obvious that the ZF method
shows poor performance with a maximum BER and fails to
outperform other methods. At the same time, the LS method
depicts somewhat better result over ZF with a lower BER
and begins to reduce certainly. It is noted that MMSE
maintains to achieve better results compared to other
methods with an increasing number of antennas from 2*2 to
4*4 under the speed of 300km/hr.
4*4 antennas
From the simulation results, it is inferred that the results
of MIMO-OFDM system with equalization increases the
performance due to the reduction of ISI. The better
improvements in the results are also due to the effective use
of time domain and frequency domain adaptively in an
appropriate way. It can also be inferred that the increase in
number of transmitting and receiving antennas and increase
in number of pilot subcarriers improve the performance due
to the diversity and multiplexing gain.
V. CONCLUSION
For adaptive resource allocation using the CSI among the
transmitter and receiver side, in this paper, we have
presented a hybrid adaptive channel estimation technique
using the integration of time domain and frequency domain.
At the initial stage, by default, the time domain based
channel estimation is good. In case of better channel quality
and low BER for a predefined time period, then a swapping
of frequency domain from time domain takes place and vice
versa. For the effective investigation on the performance of
the proposed method, an extensive set of experimentation
takes place under varying number of antennas, pilots, SC
and mobility speed of 100-300km/hr. The better
improvements in the results are also due to the effective use
of time domain and frequency domain adaptively in an
appropriate way. It can also be inferred that the increase in
number of antennas and increase in number of pilot
subcarriers increase the performance due to the diversity and
multiplexing gain.

110
Published By:
REFERENCES
1. S. T. Chung and A. J. Goldsmith, “Degrees of freedom in
adaptive modulation: a unified view,” IEEE Trans.
Commun., vol. 49, no. 9, pp.1561–1571, 2001.
2. J. Goldsmith and S.-G. Chua, “Adaptive coded
modulation for fading channels,” IEEE Trans. Commun.,
vol. 46, no. 5, pp. 595–602, 1998.
3. T. Keller and L. Hanzo, “Adaptive multicarrier
modulation: a convenient framework for time-frequency
processing in wireless communications,”Proc. IEEE, vol.
88, no. 5, pp. 611–640, 2000.
4. Svensson, “An introduction to adaptive QAM
modulation schemes for known and predicted channels,”
Proc. IEEE, vol. 95, no. 12, pp. 2322–2336, Dec. 2007.
5. M.-S. Alouini, X. Tang, and A. J. Goldsmith, “An
adaptive modulation scheme for simultaneous voice and
data transmission over fading channels,” IEEE J. Sel.
Areas Commun., vol. 17, no. 5, pp. 837–850,1999.
6. G. G. Raleigh and J. M. Cioffi, “Spatio-temporal coding
for wireless communication,” IEEE Trans. Commun.,
vol. 46, no. 3, pp. 357–366, 1998.
7. A. Goldsmith, Wireless Communications. Cambridge
University Press, 2005.
8. S. T. Chung and A. Goldsmith, “Adaptive multicarrier
modulation for wireless systems,” in Proc. 2000
Asilomar Conf. Signals, Syst. Comput., vol. 2, pp. 1603–
1607.
9. A. Olfat and M. Shikh-Bahaei, “Optimum power and rate
adaptation for MQAM in Rayleigh flat fading with
imperfect channel estimation,” IEEE Trans. Veh.
Technol., vol. 57, no. 4, pp. 2622–2627, 2008.
10. “Optimum power and rate adaptation with imperfect
channel estimation for MQAM in Rayleigh flat fading
channel,” in Proc. 2005Veh. Technol. Conf. – Fall, vol.
4, pp. 2468–2471.
11. X. Cai and G. B. Giannakis, “Adaptive PSAM
accounting for channel estimation and prediction errors,”
IEEE Trans. Wireless Commun., vol. 4, no. 1, pp. 246–
256, 2005.
12. S. Ye, R. S. Blum, and L. J. Cimini, “Adaptive OFDM
systems with imperfect channel state information,” IEEE
Trans. Wireless Commun., vol. 5, no. 11, pp. 3255–3265,
2006.
13. S. Falahati, A. Svensson, T. Ekman, and M. Sternad,
“Adaptive modulation systems for predicted wireless
channels,” IEEE Trans. Commun., vol. 52, no. 2, pp.
307–316, 2004.
14. M. Karami, A. Olfat, and N. C. Beaulieu, “Pilot symbol
assisted adaptive modulation for OFDM systems with
imperfect channel state information,” in Proc. 2010
IEEE Global Telecommun. Conf., pp. 1–6.
15. J. K. Cavers, “An analysis of pilot symbol assisted
modulation for Rayleigh fading channels,” IEEE Trans.
Veh. Technol., vol. 40, no. 4, pp. 686–693, 1991.
16. Y. Chen and N. C. Beaulieu, “Optimum pilot symbol
assisted modulation,” IEEE Trans. Commun., vol. 55, no.
8, pp. 1536–1546, 2007.
17. S. Sampei and T. Sunaga, “Rayleigh fading
compensation for QAM inland mobile radio
communications,” IEEE Trans. Veh. Technol., vol.
42,no. 2, pp. 137–147, 1993.
18. J.H. Kotecha and A.M. Sayeed, “Transmit signal design
for optimal estimation of correlated MIMO channels,”
IEEE Transactions on Signal Processing, Vol. 52, pp.
546-557, Feb. 2004.
19. C. Chuah, D.N.C. Tse, J.M. Kahn, and R.A. Valenzuela,
“Capacity scaling in MIMO wireless systems under
correlated fading,” IEEE Trans. on Information Theory,
Vol. 48., pp. 637-650, 2002.

Hybrid Adaptive Channel Estimation Technique in Time and Frequency Domain for MIMO OFDM Systems

More Related Content

What's hot (20)

Similar to Hybrid Adaptive Channel Estimation Technique in Time and Frequency Domain for MIMO OFDM Systems (20)

More from Tamilarasan N (15)

Recently uploaded (20)

Hybrid Adaptive Channel Estimation Technique in Time and Frequency Domain for MIMO OFDM Systems