Abstract
This paper proposes a novel channel estimation and prediction framework for millimeter-wave (mmWave) communications. For time-invariant channels, we develop the fast Fourier transform - gradient descent (FFT-GD) algorithm, leveraging the channel sparsity to efficiently estimate the parameters, including the angles-of-arrival (AoAs) and the complex gains. To enhance estimation accuracy, we propose the data-assisted FFT-GD (DAFG) algorithm, which jointly performs data detection and channel estimation within an expectation-maximization (EM) framework. For time-variant channels, we innovate with a Cramér-Rao bound Kalman filter (CRB-KF)-based DAFG algorithm, incorporating an autoregressive (AR) model to exploit temporal correlations of the parameters for channel prediction. This approach effectively addresses the nonlinearity challenge associated with mmWave channel parameters, resulting in improved estimation accuracy. Simulation results show that both FFT-GD and DAFG methods approach the CRB, and the CRB-KF-based DAFG maintains channel estimation accuracy across multiple frames.
Funding Information
  • National Natural Science Foundation of China (62471145)

This publication has 48 references indexed in Scilit: