Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 17

Channel estimation based on

neural network in space time


block coded
MIMOOFDM system
By Girijashankar Sahoo

Content

Objective
Introduction
MIMO OFDM
Multilayered perceptron neural
networks
LevenbergMarquardt algorithm
Simulation
Conclusion

Objective
Estimate channel parameter in OFDM
MIMO system by using feed-forward
multilayered perceptron (MLP) neural
network trained with the Levenberg
Marquardt algorithm.

Introduction
MIMO OFDM system

Transmitter

At the transmitter, serial data stream


is mapped by considering modulation
type. And pilot signals are inserted to
get CIRs, then data blocks [X(2k),
X(2k + 1)] are encoded by space
time block

OFDM symbol after applying IFT

Receiver
Received signal

hij(n) channel impulse response and


nj(n) is additive white Gaussian noise
with zero mean and variance.
Received FFT output

Transmitted data received

Multilayered perceptron neural


networks
Neural network has adaptive weights
capability with learning algorithm.In
this study the LevenbergMarquardt
learning algorithm is used to train
network.

The input signal parts which are


separated into real and imaginary and
weights sum that comes into nodes are
computed
by
applying
activation
function.
output at hidden layer

Activation function computed at output


layer

Neural estimator output

channel impulse response Hij(n) will


be estimated by adding each parts of
the outputs.
To adjust weights of input to hidden
layer and hidden to output layer, LM
algorithm is used

LevenbergMarquardt algorithm
In LM training, to adjust synaptic weights objective function E(w) that
will be minimized
Training steps
i. Initialize weight (w) and learning rate ()
ii. Compute objective function that is sum of squared errors between
desired and actual outputs

iii. Weight Update


W=W+W
iv. Recompute squared error using w + w. If error is smaller than
computed one in Step 2 then reduce the learning rate () 0.1 times, if
error is not reduced increase learning rate () 10 times and go to Step
3
v. Finish the training when error is less than the predetermined value

Simulation
BER analysis 2x2 system

MSE analysis 2x2 system

BER analysis 4x4 system

MSE analysis 4x4 system

Conclusion
Evaluate the performances, Channel
estimation based on neural network
in space time block coded MIMO
OFDM system and it is better,
compared with LS and LMS channel
estimation schemes.

Thank You

You might also like