异步DS-CDMA系统盲空时信道估计及干扰抑制

Suppressing MAI Further in Blind Space-Time Channel Estimation for Asynchronous DS-CDMA System

采用噪声子空间技术进行异步DS-CDMA系统盲空时信道参数估计,同时利用了多径传播和接收机同步失调特性,使用一种修改的ULV更新算法进行噪声子空间跟踪。为了抑制多址干扰(MA I),提出一种基于投影的辅助矢量(PAV)算法。仿真结果验证了该算法的有效性。

Existing methods suppress MAI （Multi-user Access Interference） quite effectively, but, in our opinion, MAI can be further suppressed. We propose using a projection-based auxiliary vector （PAV） algorithm to suppress MAI further in frequency selective Rayleigh fading channel. After investigating the code space of the multipath signals and the data vector space, we use the noise-subspace-based approach to perform blind channel parameter estimation in the asynchronous DS-CDMA system. This approach not only exploits the characteristics of multipath propagation but also the characteristics of timing-offsets which may occur in the receiver, thus facilitating the application of a blind linear filter-optimization technique to robust interference suppression. We modify the ULV updating algorithm and apply it to performing noise subspace tracking. The modified algorithm retains the advantage of Ref. 5＇s algorithm; it does not require rank estimation of the correlation matrix, and estimates directly the noise subspace without tracking the signal subspace. A reconstructed maximum ratio combining （MRC） filtering vector can be evaluated by using the outputs of the last filtering vector. Taking as an auxiliary vector, called A, the orthogonal projection of the reconstructed filtering vector onto the space spanned by the basic filtering vector and the previously derived auxiliary vector, we can form a new filtering vector by linearly combining the last filtering vector with auxiliary vector A. Simulation results show that we can suppress MAI much further than obtainable with existing methods as given in Refs. 1 through 4. The simulation results for BER（Bit Error Rate） can be summarized as follows： when signal-to-noise ratio per bit is in the range of 0 -8 dB, the suppression of MAI is not markedly better; but in the range of 8-20 dB, the improvement rapidly increases with increasing signal-to-noise ratio per bit; for 20 dB, BER for our method is at least one order of magnitude lower than that of existing methods. The simulation results for SINR （signal-to- interference-and-noise ratio） can be summarized as follows： when the symbol number is in the range of 0= 400, the suppression of MAI is not markedly better; but in the range of 400-2000, the improvement increases with increasing symbol number; when symbol number is 2000, the improvement is as much as 20 dB.