A new reduced complexity ML detection scheme for MIMO systems - IEEE Journals & Magazine

As a nonprofit organization, IEEE is the world's largest technical specialty organization dedicated to advancing technology for the benefit of mankind.

© Copyright 2018 IEEE - All rights reserved. By using this website, you agree to these terms of service.

For multiple-input multiple-output (MIMO) system, as the number or level of modulation antennas increases, the optimal maximum likelihood (ML) detection requires considerable complexity. In this paper, we propose a new algorithm which achieves ML performance while considerably reducing complexity. Based on a minimum mean square error (MMSE) criterion, the proposed scheme reduces the search space by eliminating the lower candidate symbol reliability in the data stream. Probabilistic metric for reliability is evaluated by using the likelihood function is normalized for each symbol candidate, to allow near-optimal ML detection. We also derived performance analysis to support the effectiveness of the proposed approach. Threshold parameters are introduced to balance the trade-off between complexity and performance. In addition, we propose an efficient method for generating log-likelihood ratio (LLR) values ​​that can be used in coding systems.

In this paper, we discuss three applications of QR decomposition algorithm for decoding in MIMO system. We propose a new lattice representation for spherical decoding. This new structure has the main effect of decrypting the real part and the imaginary part of each transmitted composite symbol independently of each other, enabling parallel detection. This reduces the number of calculations required by the receiver, which reduces the overall decoding complexity. On the other hand, in the other two applications, effective ML decoding algorithm based on QR matrix decomposition of the channel matrix have been proposed for quasi-orthogonal space-time block code and orthogonal space-time block code. While greatly reducing decoding complexity, performance is the best compared to traditional ML.