Abstract:
Users’ starvation for more reliability, high speed and capacity wireless communication
have caused the invention of 5G NR communication system. As we know the recent
communication technologies are designed on the basis of conventional communication
philosophies, which significantly limit additional performance perfections and that is the root of
daunting limitations. one of the important areas of the mobile communication is the wireless
channel estimation method which can significantly improve the performance of the whole system,
and particularly for 4G-systems and 5G-systems.
In this thesis we examine the baseline channel estimation methods used for orthogonal
frequency division multiplexing (OFDM) systems, such as the minimum mean square error
‘MMSE’ estimator and the least square (LS) estimator. We studied the MMSE and LS estimators’
architecture an examine their performances. And prove that the MMSE estimator performance is
better but it is computational complexity is high, in contrary the LS estimator has low complexity
with low performance.
Therefore, in this thesis we propose a different and efficient solution for channel estimation
which is based on machine learning techniques and in particular we used deep learning techniques
to overcome the performance issues associated with the traditional channel estimation baseline
methods, we assess the proposed estimator performance on basis of Long Short-Term Memory
(LSTM) and symbol error rate for 16 QAM systems for a multi-user communication system. We
also evaluate estimator computational accuracy and feasibility