Channel Estimation Based on Deep Learning for OCDM Communications

Abstract

Orthogonal chirp division multiplexing (OCDM) takes advantage of double spreading in both time and frequency domains for high rate communications. Multipath channels often deteriorate communication performance. Thus, channel estimation is preliminary for equalization to combat multipath. In OCDM systems, either independent pilot blocks are employed or extra null symbols have to be applied to accommodate inter-chirp interference by the conventional channel estimation methods. The bandwidth efficiency is reduced. In this paper, we develop a deep learning (DL) based channel estimation method for OCDM systems. Making use of the power of DL, the requirements of pilot symbols are relaxed and the bandwidth efficiency is increased. Simulation results corroborate the efficiency of the DL-based method compared with the conventional ones.