Supervised Learning of Channel Delay Spreads for Cyclic-Prefix Free OFDM Systems

Jami Hema Ganesh, Jen-Ming Wu
2021 IEEE 94th Vehicular Technology Conference


This paper studies the use of supervised learning on the channel delay spreads to exploit variable Guard Interval (GI) for Orthogonal Frequency Division Multiple Access(OFDMA) based systems. In a multi-user OFDMA system, the cyclic prefix (CP) as External Guard Interval (eGI) applies fixed length among users for synchronization within the cell.
However, the excessive length of eGI leads to spectral inefficiency for near users. The DFTs-OFDM scheme allows flexible length internal guard interval (iGI) and improves the SE of the systems by stagnating the total symbol duration. This paper studies the use of supervised learning technique to determine the length of iGI for each users. The training database is generated based on normalized delay and received power for Tapped Delay Line (TDL)models defined in 3GPP 5G NR technical release. The machine learning classification technique is used to classify the different channel delay spreads which can later be used to determine the length of iGI to the users. The numerical simulation shows that the spectral efficiency can be improved over the CP based OFDM system up to 27%.