Machine Learning-based Methods for Joint {Detection-Channel Estimation} in OFDM Systems

In this work, two machine learning (ML)-based structures for joint detection-channel estimation in OFDM systems are proposed and extensively characterized. Both ML architectures, namely Deep Neural Network (DNN) and Extreme Learning Machine (ELM), are developed to provide improved data detection performance and compared with the conventional matched filter (MF) detector equipped with the minimum mean square error (MMSE) and least square (LS) channel estimators. The bit-error-rate (BER) performance vs computational complexity trade-off is analyzed, demonstrating the superiority of the proposed DNN-OFDM and ELM-OFDM detectors methodologies. The conventional orthogonal frequency-division multiplexing (OFDM) system is a multicarrier scheme widely utilized in communication systems due to its capacity to combat frequency-selective fading in wireless channels. This work was partly supported by the National Council for Scientific and Technological Development (CNPq) of Brazil under Grants 310681/2019-7, and in part by the CAPES (Financial code 001), and by State University of Londrina (UEL), PR, Brazil.