Supervised-Learning for Symbol Detection in Time Varying Channels

Abstract

This paper presents a learning-based symbol detection method for time-varying inter-symbol-interference channels. Under the time-varying channel environment, accurate detection of the data symbols is challenging because of incomplete knowledge of instantaneous channel state information at the receiver. Inspired by an existing data-driven joint channel estimation and symbol detection method, our detection method is to adaptively learn the channel variation using a set of previously detected symbols as new training samples for the channel estimation. Using simulations, we show that the proposed online learning based symbol detection method outperforms the existing learning based symbol detection methods under moderate mobility scenarios.