An Efficient MIMO Detection Algorithm Employed in Imperfect Noise Estimation

Abstract:- An efficient multiple-input multiple-output (MIMO) detection (MD) algorithm includes novel, low-complexity, near-optimal and robust scheme is proposed in wireless communications when imperfect noise estimation is considered. By using MIMO technique, capacity increases proportionally as the number of antennas is increased, but the introduced inter-antenna interference (IAI) degrades system performance. To better mitigate IAI, we propose a two-stage procedure to achieve maximum likelihood (ML) performance while keeping at acceptable level of computational complexity. A novel two-stage procedure is proposed as follows that is suitable for either in an overdetermined or an underdetermined MIMO system. In an overdetermined system, interference cancellation is first processed at Stage-1 using sorted QR decomposition (SQRD) followed by Stage-2 that performs a genetic algorithm (GA). In terms of computational complexity, this procedure provides three significant advantages: 1) The SQRD scheme provides excellent interference cancellation so as to largely improve initial setting of GA. 2) By using QRD, fitness value evaluation of GA involves less complexity due to reducing the matrix multiplication. 3) In aspect of diversity knowledge, lately decoded sub-streams expect to have lower error probabilities by using SQRD. In this paper, each mutated gene is decoded from the various diversity gains, termed as a diversity mutation (DM) scheme. To achieve the fore-mentioned three significant advantages in an underdetermined system, on the other hand, we propose zero forcing (ZF) assisted SQRD GA-MD (ZF-SQRD GA-MD) to achieve ML performance. Beside, the proposed two-stage detection procedure is quite robust as it does not rely on noise information. Simulation results show that the proposed two-stage detection procedure can achieve a near-ML performance, but at a low-complexity level. Key-Words: multiple-input multiple-output, inter-antenna interference, maximum likelihood detector, sorted QR decomposition, genetic algorithm, underdetermined.