Data-Aided Channel Estimation for MIMO-OFDM Systems over Time-Varying Channels

Abstract

This thesis considers a channel estimation problem of multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems. In practical MIMO-OFDM systems, channel estimation error is inevitable because demodulation reference signals (DM-RS) for channel estimation are assigned only to subsets of subcarriers and OFDM symbols. To solve this problem, data-aided channel estimation methods based on expectation-maximization (EM) algorithm, Kalman filter, and a successive update approach for MIMO-OFDM systems are proposed. The common idea of the proposed methods is to update a channel estimate by exploiting detected data symbols as additional DM-RSs for improving channel estimation accuracy. The proposed method based on the EM algorithm iteratively updates channel estimates and data symbols based on an EM criterion. This method can significantly improve the accuracy of the channel estimate but requires high computational complexity that may not be affordable in practical systems. The proposed method based on the Kalman filter predicts and updates the channels in sequential order by assuming an autoregressive model for temporal channel evolution. The major advantage of this method is that it has a lower complexity than the EM-based method. This method, however, requires a prior knowledge about autoregressive order of channel evolution model and cannot jointly benefit from multiple DM-RSs assigned to different OFDM symbols. The proposed method based on the successive update approach estimates the channels in an appropriate order determined by the positions of the DM-RSs. In this method, the current channel to be estimated is first predicted based on the previously estimated channels and then combined with a channel update determined from the detected data symbols in the current channel. This method does not make any assumption on temporal channel statistics, while having moderate computational complexity. Using simulations, it is demonstrated that the proposed channel estimation methods outperform conventional channel estimation which utilizes the DM-RS symbols only. It is also demonstrated that the proposed methods show robust performance in time-varying channel environments and also in a low signal-to-noise-ratio regime.