多址干扰的存在是影响CDMA系统性能和容量的重要因素，多用户检测技术是克服多址干扰的有效途径。复杂度低、性能优良的多用户检测算法是目前国内外研究多用户检测技术的主要目标。由于盲多用户检测器不需要训练序列和干扰用户的先验知识，已成为多用户检测技术的一个重要的研究方向。 通过深入分析研究盲自适应多用户检测算法，特别是基于子空间的PASTd和OPAST盲多用户检测算法，针对这两种算法发散的现象进行了改进。PASTd算法的发散主要是由于估计出来的特征向量正交性不强，本文采用Gram-Schmidt方法使所得到的特征向量集合正交化，抑制其发散。OPAST算法的发散主要是由于实际计算中的误差，导致Hermitian正定矩阵不正定，使得特征值出现了误差，本文根据Hermitian正定矩阵的特性进行改进。仿真表明，改进后的算法在收敛性能，跟踪性能上都得到了提高。 另外，分析了基于卡尔曼滤波的盲多用户检测算法，与前述两种算法以及MOE，CMA盲多用户检测算法进行比较分析，仿真表明，基于卡尔曼滤波的盲多用户检测算法具有良好的收敛性和跟踪性，抑制多址干扰的能力较强。

The capacity and performance of a code division multiple access (CDMA) system is essentially dominated by multiple access interference (MAI). The well cancellation of MAI that degrades the capacity and performance could depend on multiuser detection technology. In recently years, the main research work may focus on how to present a high performance and low complexity algorithm. Blind multiuser detector has become an important study point as it has no needs of training sequence and the prior knowledge of interferential users. This paper has studied the blind multiuser detection based on subspace, and PASTd algorithm and OPAST algorithm .Both of the two algorithms have the same flaw, i.e.divergence. The divergence of PASTd algorithm is because of the eigenvectors estimated with weak orthogonality. This paper has adopted Gram-Schmidt method to orthogonalize the eigenvector set. The divergence of OPAST algorithm is because of the error of the actual computation, which make the positive definte matrix no positive definte. According to the character of the positive definte matrix, this paper has improved it. Simulation results show that the two improved algorithms are of great convergence and accurate tracking performance. And this paper also has studied the blind multiuser detection based Kalman filter, and compared with the two improved algorithms, the MOE algorithm and CMA algorithm by computer simulating. The simulation results show that the Kalman filter algorithm has a good capability of convergence and tracking, also the simulation results show that it has strong capability of suppressing MAI.