盲分离模型用于相关噪声的滤波问题

Blind Signal Separation Model Used for Filtering Relative Noise

在实际应用中经常需要从观测数据中提取出期望的信号 .现有的基于滤波器模型的算法中估计滤波器系数的Weiner_Hopf方程有唯一解要求信号的相关矩阵是非奇异的 ,并且要得到精确的估计结果需要的滤波器阶数很高导致计算量很大 ,不便于实时处理 .首次将盲分离模型用于线性估计问题的滤波问题 ,算法克服了上述基于滤波器模型的局限性 ,能对信号进行有效的提取 .

In practical applications, expected signals often need to be extracted from observed signals. Existed algorithms based on filter model use Weiner_Hopf equation to estimate the filter coefficients. To guarantee the unique solution, the correlation matrix of signals is assumed to be nonsingular. And also high filter order which induces to the complex computation is needed to get accurate estimation. The limitations make these algorithms are not suitable for real_time processing. In this paper, we use blind signal separation model for linear filtering for the first time. The new algorithm overcomes the limitations of the algorithms based on filter model. It can extract expected signal effectively.