Joint fast time domain channel estimation with ICI cancellation for LTE-R systems

In this paper, we consider the high-speed railway (HSR) communication bases on the long-term evolution for railway (LTE-R) platform. Since the large Doppler spread is introduced due to the movement of the train, the orthogonality of subcarriers is destroyed resulting in the inter-carrier interference (ICI). Moreover, the critical speed of transceivers moving leads to a very fast-time varying channel within an OFDM symbol, hence the existing frequency domain channel estimation (FDCE) method cannot reliably estimate the channel state information (CSI). To better track the fast-time varying channel, we propose a new framework for LTE-R systems. First, we modify the WINNER II channel model and the D2a propagation scenario to approximate the multipath fading and high Doppler spread in high mobility scenarios. Second, we use a novel pilot structure-based time domain channel estimation (TDCE), helping to track the channel variations for each channel path separately. The CSI at data subcarriers is estimated by using different conventional interpolation methods to reconstruct the ICI channel matrix in the frequency domain. Different from the linear channel model, the channel time-variations of current OFDM data symbol are predicted from the third-order polynomial function, which is reconstructed base on CIR at the pilot position by using several constitutive OFDM symbols. The cubic Hermite function is chosen by considering its simplicity and efficiency. Lastly, we propose the deep neural network (DNN) based CE in LTE-R systems, which helps to further improve the accuracy of estimated channel information in the previous state, especially at low signal-to-noise (SNR) level. The simulation results show that our proposed CE method for each channel path agrees well with the theoretical derivation and the simulation. The system performance with the proposed framework is significantly improved in comparison with state-of-the-art methods. (C) 2021 Elsevier B.V. All rights reserved.