Machine Learning-Driven Channel Estimation for OFDM Systems in Frequency-Selective Multipath Rayleigh Fading Channels

Orthogonal Frequency Division Multiplexing (OFDM) is highly regarded in wireless communication systems due to its high data rate efficiency, optimal bandwidth utilization, and inherent robustness against propagation-induced fading. Channel estimation (CE) plays a critical role in enhancing the performance of OFDM systems, particularly in the context of fifth-generation (5G) wireless communication. This study explores the potential of machine learning-based techniques to address the challenges of frequency-selective multipath Rayleigh fading channels. Practical channel models, including Pedestrian-B, Vehicular-A, and Urban environments, are used to generate datasets representing wireless communication scenarios. Linear regression and polynomial regression are applied to estimate channel responses, with performance assessed using metrics such as R-Squared Error and Root Mean Square Error (RMSE). Results indicate that polynomial regression performs effectively for the all the models. The findings highlight the potential of machine learning techniques to improve CE accuracy, providing a scalable solution for next-generation communication systems.