3rd International Conference on Engineering and Applied Natural Sciences, Konya, Turkey, 14 - 17 January 2023, pp.399-404
Orthogonal frequency-division multiplexing (OFDM) is commonly used in wireless
communication systems due to its high performance over frequency selective fading channels and also the
effect on reducing the inter-symbol interference (ISI). This paper investigates the performance of deep
learning-based channel estimation for OFDM systems over Weibull fading channels. In wireless
communication systems, the transmitted signal is exposed to various obstacles and fading occurs. Channel
state information (CSI) is needed to correct the errors of the transmitted signal. To obtain the CSI,
channel estimation methods such as least square (LS) and minimum mean square error (MMSE) are
widely used. Deep learning-based channel estimation can have a higher symbol error rate (SER)
performance than LS and MMSE methods. In addition, classical channel estimation methods such as the
MMSE method have complexity. Deep learning-based channel estimation can reduce the complexity of
estimation methods. In this study, the performances of the proposed method and classic channel
estimation methods in different channel conditions were compared by using computer simulations. It has
been observed that deep learning-based channel estimation has higher performance than classic methods
in different channel conditions with its ability to learn and adapt. Simulation results show that deep
learning is a promising and high potential tool for channel estimation in OFDM systems.