Low‐complexity gated recurrent unit‐based channel estimation for space–frequency block coding multiple‐input multiple‐output orthogonal frequency‐division multiplexing in high‐mobility tapped‐delay‐line‐D channels

Abstract

Reliable operation in space–frequency block coding (SFBC) multiple‐input multiple‐output orthogonal frequency‐division multiplexing (MIMO‐OFDM) systems requires accurate channel estimation, which is challenging in high‐mobility scenarios owing to rapid channel variations. Traditional pilot‐assisted estimators, such as least‐squares (LS) and linear minimum mean‐square error (LMMSE), suffer severe performance loss under these conditions. Long short‐term memory (LSTM) networks and other deep‐learning‐based methods are more accurate but computationally expensive. This study proposes a low‐complexity gated recurrent unit (GRU)‐based channel estimator that reduces computational overhead while maintaining high performance. Simulation results over tapped‐delay‐line‐D (TDL‐D) high‐mobility channels demonstrate that the proposed GRU‐ and conventional LSTM‐based estimators outperform LS and LMMSE over the primary evaluated carrier‐to‐noise ratio (CNR) range of 0–20 dB in terms of mean‐square error and bit error rate. Whereas LSTM slightly outperforms the proposed GRU at extremely low CNRs, the proposed GRU achieves superior overall performance, particularly at high CNRs, where LSTM may saturate.