양자화 및 압축된 신호로부터 유한 알파벳 희소 신호 복구에 관한 알고리즘 연구

Abstract

In this thesis, a problem of detecting finite-alphabet sparse signals from noisy and coarsely-quantized measurements is considered. To solve this problem, a greedy sparse signal detection algorithm referred to as Bayesian matching pursuit (BMP) is proposed. The key idea of BMP is to identify the non-zero elements of a sparse signal that produce the largest a posterior probabilities in an iterative fashion. In this procedure, the posterior distribution following massive mixture of Gaussian distribution is efficiently approximated by a simple Gaussian distribution. The BMP is a general framework that can provide multiple variations. For single measurement vector (SMV) problems, a generalization of BMP called Bayesian multipath matching pursuit (BMMP) is also presented. In addition, generalizations of the BMP for multiple measurement vector (MMV) problems are discussed. From simulations, it is shown that the BMP and its variations can outperform the existing sparse signal reconstruction algorithms in terms of frame error rates (FERs) even with a significantly reduced computational complexity. Further, a possible application of the proposed algorithm is discussed in the context of wide-band multiple-input multiple-output (MIMO) communication systems using low-resolution analog-to-digital converters.