부산대학교 도서관

This paper presents a novel Time Domain (TD) based Multiple-Input Multiple Output (MIMO) implementation of the Expectation-Maximization (EM) algorithm. The algorithm is able to achieve Maximum Likelihood (ML) performance with reduced complexity. Results show how performance superior to the Recursive Least Square (RLS) algorithm, in terms of both convergence and residual Mean Squared Error (MSE), can be achieved with comparable complexity. The channel estimation technique is combined with a new TD based MIMO loop interference cancellation system for wireless relays. Performance results for a 2x2 Downlink (DL) Long Term Evolution (LTE) system are provided for both single-stream transmission with Maximal Ratio Combining (MRC) and dual-stream transmission with Spatial Multiplexing (SM). Results are shown at both low mobility and high mobility, with our EM algorithm showing up to a 2dB improvement in the Error Vector Magnitude (EVM) in the MRC case at 60kmph compared to the RLS.