MIMO Receiver Structures with Integrated Channel Estimation and Tracking

Abstract

This thesis looks at the problem of channel estimation and equalization in a multiple-input multiple-output (MIMO) dispersive fading environments. Two classes of MIMO receiver structure are proposed with integrated channel estimation and tracking. One is a symbol-by-symbol based receiver using a MIMO minimum mean square error (MMSE) decision feedback equalizer (DFE), and the other is a sequence-based receiver using a partitioned Viterbi algorithm (PVA) which approaches the performance of maximum likelihood sequence estimation (MLSE). A MIMO channel estimator capable of tracking the time and frequency selective channel impulse responses, known as the vector generalized recursive least squares (VGRLS) algorithm, is developed. It has comparable performance and a similar level of complexity as the optimum Kalman filter. However, it does not require any knowledge of the channel statistics to operate and as such it can be employed in a Rician fading channel readily. A reduced complexity form of the estimator, known as the vector generalized least mean squares (VGLMS) algorithm, is also developed. This is achieved by replacing the online recursive computation of the VGRLS algorithm's 'intermediate' Riccatti matrix with an offline pre-computed matrix. This reduces the complexity of the algorithm by an order of a magnitude, but at the expense of degraded performance. The estimators are integrated with the above-mentioned equalizers in a decision directed mode to form a receiver structure that can operate in continuously time-varying fading channels. Due to decision delays, the outputs from the equalizer are delayed and this then produces 'delayed' channel estimates. A simple polynomial-based channel prediction module is employed to provide up-to-date channel estimates required by the equalizers. However, simulation results show that the channel prediction module may be omitted for a very slowly fading channel where the channel responses do not vary much. In the case of the PVA- receiver, the zero-delay tentative decisions are used as feedback to the channel estimators with negligible loss.