OFDM CHANNEL ESTIMATION USING BAYESIAN REGULARIZED DEEP LEARNING

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) has clearly emerged as one of the most potent enablers for high data rate and high spectral efficiency communication systems. Most wireless communication systems consisting of large number of users sharing a common channel with limitations in bandwidth opt for OFDM transmission. However, due to the frequency selective nature of wireless channels, OFDM often faces degradation in the Bit Error Rate (BER) and hence Quality of Service. This paper proposes a channel estimation mechanism for OFDM based on Bayesian Regularized ANN. It has been shown that the proposed approach attain low BER and lesser number of iterations for training compared to conventional systems. The deep learning algorithm used is the Bayesian Regularization which is an effective algorithm for analyzing large and complex patterns in data. The performance metrics are the BER, mean square error (MSE) and number of epochs.