Low‐complexity soft‐output MIMO uplink detection for large systems iterative detection and decoding

Abstract

Iterative detection and decoding techniques, based on the turbo principle, have been proposed in the literature to achieve near‐capacity on both single‐ and multiple‐antenna fading communication systems. In the multiple‐input–multiple‐output case, when a very large system is considered, one major issue is the overall system complexity. Here, we propose an iterative detection and decoding scheme for the uplink of coded multiple‐input–multiple‐output systems that replaces the exponential complexity maximum a posteriori detector by an approximate maximum a posteriori low‐complexity detector that uses a simple hard‐output inner detector at the first iteration followed by parallel interference cancelation and filtering matched to the channel coefficient matrix that is estimated with the aid of orthogonal pilot transmission. From the second iteration onward, the inner hard‐output detector is replaced by the results from the previous iteration. Simulation results demonstrate that the proposed strategy presents good bit‐error‐rate performance with low computational complexity, measured by the average number of floating‐point operations per message bit.

A Seysen's algorithm–based incremental lattice reduction

“…As the columns of H LR are roughly orthogonal, the noise amplified by it is much less than the one in Equation (2). Therefore, the reliability of quantisation on symbolz eq is higher than on the symbol s eq , and by using (7),z eq can be expressed as…”

“…Hence, suboptimal linear and nonlinear detectors with low complexity are often employed. [5][6][7] However, these schemes cannot obtain the full receive diversity and experience a significant performance loss. Observing this, the lattice reduction (LR) technique is introduced to improve their performance.…”

Detecção Desacoplada De Sinais No Enlace Reverso De Sistemas Mu-Mimo Com Modulação Espacial

“…. , N ), (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18) ondeȞ n = Q n H n ∈ C Mt n ×Mt n é a matriz de canal equivalente da n-ésima classe de usuário depois da DSD e n n = Q n N m=1,m =n H m s m + Q n n é o vetor de ruído equivalente. Ressalte-se que como Q n tem linhas ortonormais, o vetor de ruido n = Q n n permanece branco com matriz autocorrelação K n = E[n n H ] = σ 2 n I Mt n É importante ressaltar que os processos de detecção que envolvam o desacoplamento do sinal em grupos (ou classes) de usuários estabelecem um importante compromisso entre qualidade da detecção e complexidade do receptor.…”

“…A decomposição em valores singulares (SVD) permite calcular as bases para o espaço nulo à esquerda deH n . Aplicando a decomposição emH n temos:H n =Ũ nΣnṼ H n , (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) ondeΣ n ∈ C Nr×(Mt−Mt n ) é uma matriz diagonal retangular com os valores singulares deH n na diagonal,Ũ n ∈ C Nr×Nr eṼ H n ∈ C (Mt−Mt n )×(Mt−Mt n ) são matrizes unitárias. Se r n é o posto deH n , r n =rank(H n ) ≤ M t − M tn a decomposição SVD pode ser expressa da forma:…”

“…O desempenho dos cenários apresentados nas Figuras 4.4 e 4.7 é avaliado de uma maneira mais real incluindo efeitos de propagação e correlação. Isto é feito seguindo o cenário de canal descrito em [15].…”