Abstract:
In this paper, in order to estimate channel impulse responses in orthogonal frequency
division multiplexing (OFDM), we use radial basis function network that is a kind of
neural network, because of this structure is applicable of this kind of problem for its
strong approximation and learning ability. We compare the performance of channel
estimator based on radial basis function neural network with LS and MMSE
algorithm with bit error rate (BER) and mean square error (MSE) criterias. Also
Cramer Rao bound is given to evaluate the performances of estimators. Our proposal
channel estimator has better performance than LS algorithm and closer performance
to MMSE algorithm. However there is unnecessity of knowledge of channel statics
and noise information of channel when neural structures are used as a channel
estimator. Moreover after neural structures are trained, there is no need of sending
pilot tones that are used to get channel impulse responses by LS and MMSE
algorithm. As a result, system spendings are reduced.