Optimization of Performance of MIMO-OFDM system using Neural Network as Channel Estimator and Compensator

Optimization of Performance of MIMO-OFDM system using Neural Network as Channel Estimator and Compensator

Rahul Shahane

M-Tech Scholar(CSE) TGPCET ,Nagpur,India

Prof.Roshani Talmale

HOD (CSE)

TGPCET, Nagpur,India

ABSTRACT

MIMO-OFDM systems are one of the systems which have become the basis of many communication researches nowadays.

used in MIMO in same frequency band [4] which increases the capacity linearly with the number minimum of transmit and receive antennas. However it imposes a challenge that is

Combination of both stands a good possibility of being the next-

the increased complexity [5] of channel equalization (to

generation (4th generation) of mobile wireless systems. The

separate all the signal paths and to remove the changes caused

increased complexity of channel equalization is a major

by the channel) on receiving side. Because of high degree of

challenge. Wireless channels are responsible for adding Inter

non-linearity of NN, they can be effectively used to decode

Channel Interference (ICI) and Inter Signal Interference (ISI) .

symbols transmitted through difficult channels. Equalization

For removing the effect (imposed by channel) from received

and compensation of channel can be also regarded as a

signal, the receiver needs to have knowledge of CIR (Channel

classification task [6]. In particular, due to their universal

impulse response) it is usually provided by a separate channel

approximation capability, NN can form arbitrarily shaped

estimator. Here we proposed the technique of Neural

decision boundaries [7]. In recent years NN have been often

Network(NN) as channel estimator and compensator. We train the NN with different algorithms. Then length of known training

proposed for digital equalization of communication channels.

sequence varied and observations are made. Further the whole

The research proposes a technique, which based on artificial

system is compared with the traditional algorithms and

neural networks, carries out (MIMO-OFDM) channel

observations are made to show the effectiveness of NN based channel estimator and compensator for MIMO-OFDM system.

estimation and compensation.

Estimation of channel is calculated in terms of synaptic weights and bias values of neural network, whereby, different

Keywords-MIMO-OFDM; channel estimation; neural network

training

algorithms have

been

analyzed

to calculate

those

weight and bias values.

This

research also attempts

the

1. INTRODUCTION

   usefulness

   of neural

   network

   based channel

   estimator by

   The targets of communication systems are to provide

   comparing results obtained by use of different NN-training

   services that include video, voice and data with high speed and

   algorithms. To ascertain the flexibility and performance of the

   reliability. MIMO (Multiple Input Multiple Output) and

   proposed technique; length of the training sequence has been

   OFDM (Orthogonal Frequency Division multiplexing)

   varied over a reasonable range and the result has been

   together give rise to 4th generation technology. MIMO

   observed. Then, the results obtained by using different

   communication system is a technology to achieve the targets

   algorithms for training NN have been compared with each

   of high data rates by taking advantage of multipath signals [1].

   other and against the traditional least square algorithm for

   OFDM provides resistance to ISI (Inter symbol Interference)

   channel estimation.

   and ICI (Inter carrier interference) [2]. With both technologies (MIMO & OFDM) bringing a good possibility of being the

2. PROPOSED DESIGN

   next-generation (4th generation) of fixed and mobile wireless

   Following are the Phases of Implementation of Proposed

   systems [3]. The wireless channels are multipath fading

   Design

   channels, causing ISI (inter symbol interference), whereby, for each path there is an independent path delay, independent path gain (or loss) and independent path phase shift. So to remove channel effect from the received signal, the receiver needs to

   Phase I

   As shown in below fig.1 the 2*2 MIMO channel is chosen from Matlab Simulink. Random signals are passed through

   have knowledge of CIR (Channel impulse response), it is

   MIMO channel. BPSK modulation is used for modulating the

   usually provided by a separate channel estimator. Channel

   data. BPSK modulation technique is used because it is very

   estimation is based on the known sequence of bits, which is

   efficient modulation technique.

   unique for each transmitter and is transmitted in each

   The mathematical model for 2*2 MIMO system is given

   transmission burst. Multiple transmit and receive antennas are by

   In MIMO systems, a transmitter sends multiple streams by multiple transmit antennas. The transmit streams go through a

   The mathematical model for OFDM is given as follows. The problem of inter channel interference(ICI) existing in

   matrix channel which consists of all paths between the

   an OFDM system under a time -varying channel is given and

   transmit antennas at the transmitter and receive ant t r t the r

   properties are discussed. fig 1 shows a discrete-time baseband

   ennas at he receive . Then, the receiver gets eceived

   sig and

   equivalent model for OFDM system. Input bits are encoded

   nal vectors by the multiple receive antennas decodes th t form

   into a symbol Xm. and N symbols are sent to serial to parallel

   e received signal vectors in o the original in

   narrowband flat fading MIMO system is modeled as

   y=Hx+n ………(1)

   ation. A

   converter(S/P). The inverse fast Fourier transform(IFFT) is

   then applied. The nth output of the IFFT Xn can be expressed as follows.

   where

   y and

   x are

   the

   receive

   and

   transmit

   vectors,

   respectively, and H and n are the channel matrix and the noise vector, respectively.

   Referring

   to information

   theory,

   the

   ergodic

   channel

   capacity of MIMO systems where both the transmitter and the receiver have perfect instantaneous channel state information is[13]

   ………..(6)

   Before

   the

   parallel

   to serial

   converter(P/S),

   the

   cyclic

   prefix is

   added

   to avoided

   inter-block

   interference

   and

   ….(2)

   preserve

   orthogonality

   between

   subchannels.

   Generally

   the

   length

   of the

   cyclic

   prefix is

   chosen

   such

   that

   the

   guard

   where denotes Hermitian transpose and is the ratio

   interval is longer than or equal to the delay spread of the

   between transmit power and noise power (i.e., transmit SNR).

   channel. the cyclic prefix is ignored for simplicity in this

   The optimal signal covariance is achieved

   analysis, however. By assuming that the channel consist of L

   through singular value decomposition of the channel matrix

   multipath components, and changes at every sample, the

   and an optimal diagonal power allocation matrix

   output of the channel can be given by

   . The optimal power

   allocation is achieved through waterfilling,[14] that is

   where

   are

   the

   …(3)

   diagonal

   ………(7)

   elements of D,

   is selected such that

   is zero if its argument is negative, and

   .

   where hn;l and wn represent the channel impulse response

   (CIR) of lth path and additive white Gaussian noise (AWGN)

   If the transmitter has only statistical channel state inf n g e a

   at time n, respectively. From (2.1), yn can be written as

   ormatio , then the er odic chann l capacity will decre se as

   the signal covariance can only be optimized in terms of the

   average mutual information as

   ..

   ..

   …(4)

   The spatial correlation of the channel have a strong impact on the ergodic channel capacity with statistical information.

   If the transmitter has no channel state information it can select the signal covariance Q to maximize channel capacity

   under

   worst-case

   statistics,

   which

   means

   and

   …………(8)

   accordingly

   …..(5)

   Depending on the statistical properties of the channel, the

   ergodic capacity is no greater than times larger

   …………(9)

   than that of a SISO system. System Model under a Time- Varying Channel.

   where hn;l and wn represent the channel impulse response

   These

   modulated

   signals

   are

   now

   given

   to the

   OFDM

   (CIR) of lth path and additive white Gaussian noise (AWGN)

   transmitter. The known Pilot symbols are also added to the OFDM transmitter.

   at time n, respectively. From (6.7), yn can be written as

   ……(14)

   …………………………….(10)

   ………………..(15)

   Here, Wm, ®m, and ¯m represent the Fourier transform of wn, the multiplicative distortion of a desired subchannel m, and the interchannel interference caused by a time-varying channel, respectively. Note that ®m is the average frequency

   response of the CIR over one OFDM symbol period. If the

   where H n (m) is the Fourier transform of the channel impulse

   channel

   is timeinvariant, in

   other

   words,

   H(k) n is

   not a

   response at time n.

   function

   of n,

   then

   ®m simply

   becomes

   the

   frequency

   Then, yn can be rewritten as

   …………. (11)

   response of the CIR, as expected.

   We can express (6.7) in a compact vector-matrix form as

   …………………………………….. (16)

   After removing the cyclic prefix, the demodulated symbol Ym at the receiver is

   obtained by applying the fast Fourier transform (FFT) so that

   and

   From (6.5), Ym can be written as

   …………………… (12)

   Here, Hm;k is defined as

   ……..(17)

   …………………………………. (13)

   ……………… (18)

   In an OFDM system over a time-varying channel, the inter channel interference can be characterized by the normalized Doppler frequency fdT where fd is the maximum Doppler frequency and T is the time duration of one OFDM symbol. Hence we can think of the normalized Doppler frequency as a

   Where

   maximum cycle change of the time-varying channel per

   OFDM symbol duration in a statistical sense.

   ¯ms

   or off-diagonal elements of H in

   represent the inter

   channel interference (ICI) caused by the time-varying nature of the channel. In a time invariant channel, one can easily see that ¯m is zero, or H becomes a diagonal matrix, due to the

   orthogonality of the multicarrier basis waveforms. In a slowly time-varying channel, i.e., the normalized Doppler frequency fdT is small, we can assume Efj¯mj2g ¼ 0. On the other hand, when the normalized Doppler frequency is high, the power of the ICI cannot be ignored, and the power of the desired signal is reduced.

   Phase II

   The OFDM receiver receives the signals and this data is now

   …..(22)

   ……..(23)

   R represents the output of Layer#2 (the output layer).

   ….(24)

   r1 is the output received from first neuron of layer#2 (estimate of the real part of transmitted signal) and r2 is the output received from second neuron of layer#2 (estimate of the Imaginary part of transmitted signal).

   used as

   the

   Training

   dataset

   for

   the

   Neural

   Network.The

   The Neural Network

   is trained using this dataset. We use

   training dataset is shown in fig 2. Neural network is provided with the input value and the target value for output, training algorithm calculate the weights & bias values of the network

   Levenberg-Marquradt algorithm for training of NN .

   After training the MATLAB creates a trained NN block separately .The trained NN box is shown in fig 3. This trained

   using

   input

   and

   target

   values

   provided.

   Received

   data

   block of NN can be used anywhere in the system.

   sequences are passed through the network and straight

   Phase III

   computations are made to calculate the estimate of transmitted signal. The output of layer#1 is computed as

   Z = purelin {(W{1} I) + B{1} }

   Where I is the input to the neural network and is providedwith the received signals.

   ……………..(19)

   Z represents the output of layer #1.

   Now we apply this trained NN block to our MIMO-OFDM system.After applying trained NN block to previous system as shown in fig 1. our system looks as shown in fig below

   Phase IV

   The whole system is now complete as shown in fig.4.Now vary the Signal-To-Noise (SNR)ratio using AWGN channel and observe the changes of Bit-Error-Rate (BER).

3. Simulation of MODEL

   Simulate the model from 0-25 dB SNR and make

   ……(20)

   ……..(21)

   observations. And Graphs are plotted for SNR vs BER.

   Fig 1: MIMO-OFDM model with BPSK modulation.

   Fig 2: Dataset from MIMO-OFDM system. Fig 3: Trained Neural Network Block.

   Fig 4: MIMO-OFDM system using Neural Network as Channel Estimator and Compensator

4. RESULT AND CONCLUSION.

   performance of MIMO-OFDM system for varying value of SNR in AWGN channel and its corresponding BER for the

   system. The graph shows two channels. The channel 1 is

   The model is simulated .The simulation takes place for

   without estimation and channel 2 is with NN estimation. The

   SNR 0-25 and the observation is made for varying BER.The

   significant difference is observed in both channel. It is

   Graphs are plotted for SNR vs BER for different values of

   observed that the performance of MIMO-OFDM is improved

   SNR. Following are the graphs for whole system. Graph

   when the NN as channel estimator is used.

   shows the relation between SNR vs BER. It shows the

   This paper presents a NN-based technique to estimate and compensate channel effect for MIMO-OFDM communication

   systems.

   Through

   experimentation,

   algorithm

   Levenberg-

   Marqurdt (LM) has been tested to train neural networks, it has been established that LM can effectively train neural networks

   1. Xihong CHEN, Qiang LIU, Maokai HU Missile Institute of AFEU Sanyuan, Shanxi, China ; Performance Analysis of MIMO-OFDM Systems on

      and track channel effect, which measures performance in

      Nakagami-m Fading Channels; IEEE Trans.on Info.Theory, vol.49, no.10, pp.

      terms of BER for a range of SNR values. The curves drawn as

      2363-2371, Oct.2010.

      result of the simulation; corroborate the aforementioned fact

      [5]

      Ashraf A. Tahat School of Electrical Engineering Princess Sumaya

      that LM provides better results.

      University

      for

      Technology

      Amman,

      Jordan;

      A Matrix

      Scheme to

      Fig 5: SNR vs BER in dB.

      Channel 1

      Extrapolation and Interpolation for a 4G MIMO OFDM System; IEEE Transactions on Signal Processing, vol. 51, pp June 2009

      Q. Li, G. Li, W. Lee, M. Lee, D. Mazzarese, B.Clerckx, and Z. Li, "MIMO techniques in WiMAX and L TE: a feature overview," IEEE Commun. Magazine, vol. 48, May. 2010.

      1. 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.

      2. A. Kalayciogle and H. G. Ilk, "A robust threshold for iterative channel estimation in OFDM systems," Radio Engineering journal, vol. 19, no. 1, pp. 32-38, Apr. 2010.

      3. R. C. Alvarez, R. Parra-Michel, A. G. o. Lugo,and J. K. Tugnait, "Enhanced channel estimation using superimposed training based on universal basis expansion," IEEE Trans. on Signal Process., vol. 57, no. 3, pp. 1217- 1222, Mar.2009.

      4. A. Kalayciogle and H. G. Ilk, "A robust threshold for iterative channel estimation in OFDM systems," Radio Engineering journal, vol. 19, no. 1, pp. 32-38, Apr. 2010.

   Channel 2

   [11]A. Omri , R. Hamila , M. Hasna

   , R. Bouallegue

   and H. Chamkhia

   Performance of

   MIMO-OFDM

   with NN

   Estimation of highly selective channels for downlink LTE MIMO-OFDM

   channel estimator

   system by a robust neural network Journal of Ubiquitous Systems and

   Performance of MIMO-OFDM without

   Pervasive Networks Volume 2, No. 1 (2011) pp. 31-38

   channel estimator

5. FUTURE SCOPE

   Length of known training sequences in the proposed system is

   [12]Jae Won Kang, Younghoon Whang, Hyung Yeol Lee, and Kwang Soon Kim Optimal Pilot Sequence Design for Multi-Cell MIMO-OFDM Systems IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO. 10, OCTOBER 2011

   a constraint over the speed of communication, so less the

   [13]G. Joselin Retna Kumar Low Complexity Algorithm For Channel

   length of training sequence more will be the speed. In future work is to reduce length of training sequence by maintaining

   Estimation of UWB MIMO-OFDM Wireless Fading Channels 978-1-4244- 9005-9/10 ©2010 IEEE

   the performance of system. Reducing the computational power

   demand can be possibly achieved by having pre-calculated

   1. Xiuyan Zhang , Yajie Su , Guobin Tao Signal Detection Technology Research of MIMO-OFDM System 2010 3rd International Congress on Image

      synaptic weights for different environments. Unsupervised

      and Signal Processing (CISP2010)

      learning of neural network may get the work done with less computational power.

   2. Wikipedia about MIMO and OFDM , last accessed 12 August 2013

6. REFERENCES

1. K. Elangovan

   Dept. of Computer Science and

   Engineering PRIST

   University Thanjavur, Comparative study on the Channel Estimation for OFDM system using LMS, NLMS and RLS Algorithms, Indian Proceedings of the International Conference on Pattern Recognition, Informatics and Medical Engineering, March 21-23, 2012.

2. Md.

   Masud

   Rana

   Department of

   Electronics

   and

   Communication

   Engineering Khulna University of Engineering and Technology, Bangladesh. Performance Comparison of LMS and RLS Channel Estimation Algorithms

   for

   4G MIMO

   OFDM

   Systems

   , Proceedings

   of 14th

   International

   Conference on Computer and Information Technology December, 2011,

3. Yigit, Halil Department of Electronics and Computer Education, Kocaeli University Izmit, Kocaeli, 41380, Turkey Kavak, Adnan ; Adaptation using neural network in frequency selective MIMO-OFDM systems; Wireless Pervasive Computing (ISWPC), 2010 5th IEEE InternationalSymposium.