MMSE channel factorization for near maximum-likelihood detection of MIMO signals

• Main Authors: Chih-Cheng Kuo, Wern-Ho Sheen, Chang-Lung Hsiao
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial neural networks; channel factorization-aided detectors; Complexity theory; Detectors; Lattices; MIMO; minimum mean square error; ML detection; Optimized production technology; Transforms
• ISSN: 2157-8095; 2157-8117
• DOI: 10.1109/ISIT.2008.4595052

This paper investigates channel factorization-aided detection for MIMO (multiple-input, multiple-output) spatial multiplexing signals. In essence, this technique transforms, through channel factorization, the MIMO system into an equivalent one with a better-conditioned channel that is more suitable for low-complexity detection. The detected signal in the transform domain is then transformed back to the original domain and decision is made to the transmitted signal. Channel factorization-aided detector has recently become of great interest for MIMO systems because of its ability to cope with noise enhancement incurred by an ill-conditioned channel. Lattice-reduction-aided detector (LRAD) is such an example, where channel factorization is done with the Lenstra-Lenstra-Lovasz (LLL) or Seysen lattice-basis reduction algorithm. In this paper, a new minimum mean square error (MMSE)-based channel factorization is proposed. By minimizing mean square error in the transform domain, the proposed method can approach nearly to the performance of maximum likelihood detection and outperform the LLL factorization by up to 0.5 dB, depending on the types of used low-complexity detectors and/or antennas numbers.