Semi-blind Turbo Detection for Multiple-input Multiple-output Wireless Systems

Abstract

Multiple-input multiple-output (MIMO) techniques constitute the core of future high speed wireless systems. Blind detection in MIMO wireless systems conserves bandwidth and is better suited for use in multiple user scenarios. Since, most of the modern wireless communication systems employ some form of forward error correction, therefore, we can exploit the turbo principle to further improve the performance of blind detection schemes for MIMO.

In this thesis we propose a new semi-blind turbo multiuser detector (MUD) for signal detection in a MIMO wireless system, operating on a single link with Gaussian noise. This turbo MUD (named as T-MUK) performs a sub-optimal joint detection and decoding by iteratively exchanging soft information between the detector stage, that optimizes multiuser kurtosis maximization criterion, and the decoder stage, that runs maximum aposteriori probability (MAP) algorithm. It is shown to achieve good performance at the expense of very few training symbols.

This thesis also introduces a successive interference cancellation based semi-blind MUD for multicell MIMO systems. This MUD uses T-BLAST, which is a near optimal detection structure for single link, to detect desired signals and T-MUK to detect interfering signals and improves the estimate for desired signal by using these in an iterative fashion. Numerical results indicate the advantage of using this MUD structure over T-BLAST (which treats inter-cell interference as additional noise). The semi-blind nature of T-MUK in this MUD allows us to avoid a common problem in multicell systems, that of extracting reliable channel information for interferers using training sequences.