估计算法的鲁棒性问题研究

本文针对异步DS-CDMA系统,在理论上分析了信道阶、信道头和尾对子空间盲特征波形估计算法性能的影响,并进行了计算机仿真.首先提出了一个特征波形分解模型,将特征波形分解为m阶主要部分和次要部分,得到了相应的m阶子空间算法;然后运用矩阵扰动理论,证明了特征波形分解的可行性,即m阶子空间算法的特征波形估计逼近于m阶主要部分;最后从矩阵特征值分解的稳定程度出发,推导了特征波形分解准则,该准则指出应该避免考虑次要部分,采用尽可能小的信道阶.仿真结果验证了理论分析的正确性.

On the Robustness of the Subspace Method for Blind Signature Waveform Estimation in Asynchronous DS-CDMA Systems

The effects of the channel order and tails on the performance of subspace based blind signature waveform estimation method in asynchronous DS CDMA systems are analyzed in theory,and the simulations are carried out.First,we present a signature waveform decomposition model,which means partitioning the signature waveform into the mth order significant part and the tails,and the corresponding mth order subspace method is obtained.Secondly,by using matrix perturbation theory,we prove the decomposition validation,which shows that the signature waveform estimation of the mth order subspace method is close to the mth order significant part.Finally,from the stable degree of matrix eigenvalue decomposition,the decomposition criterion is derived,which points out that we should avoid modeling the tails and adopt the channel order as small as possible.The simulation results support our theoretical analysis.