Carrier Frequency Offset And Channel Estimation For OFDM Systems Using ICA

As the core technology of 4G mobile communication, Orthogonal Frequency Division Multiplexing(OFDM) technology has many advantages, such as strong ability to resist Frequency selective fading, high Frequency band utilization, high speed data transmission, etc. Although after a long period of development, the key technologies of OFDM is gradually mature, but there are still some problems need to be improved. In this paper, by using Independent Component Analysis(ICA) technology, the frequency offset and channel estimation for OFDM systems are researched. The main content is summarized as follows:Firstly the research object of this paper OFDM is introduced, involving its basic principle, the advantages and disadvantages, the research status of carrier frequency offset and channel estimation for OFDM system. At the same time, the key technologies of OFDM system associated with this article are analyzed in detail.Secondly we introduce the technology used in this paper, Independent Component Analysis(ICA). The definition of ICA is described and the means to measure the independence of signals are also summarized. Besides two kinds of algorithms FASTICA and NLRICA are introduced briefly.And then a novel approach to estimate carrier frequency offset(CFO) of OFDM systems using ICA is proposed. Classical ICA algorithms require multiple mixed signals to perform signal separation. However, only by processing single channel OFDM data with the correlation detection technology, the new method can separate the independent components which include CFO information using ICA and then estimate the CFO of the OFDM signals. Compared with the traditional frequency offset estimation methods, this method requires no channel state information(CSI), auxiliary information and additional bandwidth, thus greatly reducing the complexity of system implementation.Finally we put forward an improved channel estimation method for zero padding(ZP) OFDM using conjugate prior. First, the channel matrix can be seen as a circulant matrix thanks to the zero padding. Then with adopting blind signal separation(BSS) technique, we only need to obtain a diagonal matrix after diagonalizing the circulant matrix by pre-multiplication the received signals by an IFFT matrix, which greatly reduces the computational complexity.Moreover, sparse priors on the diagonal matrix is added to objective function in order to improve the efficiency and accuracy of channel estimation. The proposed approach is confirmed that has lower complexity, faster convergence and higher accuracy by computer simulations, compared with the traditional algorithm by directly separating the received signals.