REGULARIZING ALGORITHMS BASED ON KALMAN FILTERING FOR M-QAM SIGNAL DETECTION IN MIMO

The aim of the article is to increase the noise immunity of signal reception with a satisfactory computational complexity of signal processing algorithms. The article discusses two regularizing recurrent algorithms for detecting quadrature amplitude modulation (M-QAM) signal in a multiple-input and multiple-output (MIMO) method and a direct transform receiver based on the Kalman filter. Kalman filtering using for soft decision detection operates at a fixed point in time and estimates symbols in iterations. The proposed detection algorithms contain a regularizing parameter. For one algorithm, the regularization parameter is selected empirically, for the other, it is found by a closed expression, which includes estimations of symbols obtained at the last step of the iterative algorithm. Hard decisions are determined by the criterion of the minimum distance between the received soft decisions and the possible values of symbols for each transmitting antenna separately. The proposed detectors are compared in terms of noise immunity (in a system without coding) with the Zero Forsing method and an algorithm that works according to the root-mean-square error (RMS) criterion. The channel is supposed to be known or it is estimated by the least squares (LS) method using first-order polynomial approximation. The article claims that regularizing detection algorithms make it possible to increase the noise immunity of signal reception relative to the Zero Forsing algorithms and RMS. In addition, this article contains an analysis of the computational complexity of the proposed recurrent algorithms. The use of the regularization parameter makes it possible to reduce the number of iterations in the detection algorithm needed to obtain the required error probability per symbol. This can reduce the number of arithmetic operations, resulting in a reduction in the computational complexity of signal processing algorithms. The proposed detection algorithms are more complicated than the Zero Forsing and RMS, but it is much simpler than the detector synthesized according to the maximum likelihood (ML) algorithm.