Block-fading channel estimation for MIMO OFDM systems via meta-learning

Abstract

In this study, we propose deep learning (DL) approach to pilot assisted channel estimation for multiple input multiple output (MIMO) orthogonal frequency multiplexing (OFDM) communication system. Conventional DL algorithms entails large number of training data and updates when deal with unseen various channels in practical. In order to reduce resources like training data and updates, we apply meta learning to deep learning based channel estimation. The performance of designed meta learning based channel estimation is evaluated under 5G standard channel models. We present numerical results on the performance of meta learning estimation and show that it can achieve the higher accuracy than other conventional estimations with just a few training samples and iterations.