Efficient detection algorithms for MMSE-SIC in MIMO systems

Abstract

In this thesis, computationally efficient detection algorithms for the minimum mean square error (MMSE) estimation with successive interference cancellation (SIC), or with generalized decision feedback equalizer(GDFE) are considered in spatial multiplexing multiple-input multipleoutput (MIMO) systems. The MMSE-SIC and the MMSE-GDFE architectures are known to achieve the ergodic capacity of MIMO fading channels. This thesis presents a new understanding of SIC and GDFE to provide further insights. This perspective shows that 1) SIC and GDFE are a special case of the two-dimensional estimation with side information (or decision feedback), and 2) SIC can be expressed in terms of the Cholesky decomposition, as GDFE is. Inspired by the latter, we propose fast algorithms based on the Cholesky decomposition for SIC and GDFE in multiple antenna systems. Although existing detection algorithms can provide an impressive reduction of computational complexity for a large $\It{M}$, the reduction is relatively small for a moderate $\It{M}$ of practical interest (i.e. $\It{M}$ $\le$ 16). The Cholesky decomposition is computationally more efficient than the QR decomposition. Specifically, the required number of multiplications for the Cholesky decomposition is O($M^3$/6) for an $\It{M}$$\times$$\It{M}$ matrix, whereas that of the QR decomposition amounts to O($4M^3$/3). Efficient and stable Cholesky decomposition based detection algorithms further reduce the computational complexity of existing fast algorithms for SIC in MIMO at fading channels. To this end, we derive special properties of the Cholesky decomposition to render the proposed algorithms more efficient computationally. A complexity comparison shows that the proposed algorithms can reduce the computational complexity of SIC and GDFE more significantly than existing fast algorithms. Finally, proposed detection algorithms for the MMSE-SIC and the MMSE-GDFE receiver can be readily extended to linear space-time co...