含噪混合数据中相关源信号的盲分离

Blind separation of correlated sources in noisy mixtures

本文在假设时间白噪声的基础上,通过对阵列施加一个较弱的旋转不变约束提出了一种针对相关源信号盲分离的方法。与现有方法相比,该方法不需要额外的源信号约束条件和噪声自相关矩阵的知识,因此具有较广的使用范围和较强的稳健性。另外,该方法较低的计算复杂度易于其工程实现。计算机仿真结果表明,本文提出的方法对于相关源信号的盲分离具有较好的效果。

Under the mild assumptions of temporally white noise and rotational invariance array, an efficient method is proposed to separate mutually correlated source signals from their instantaneous mixtures. Compared with existing methods, the proposed method does not need other constraint on the sources and the knowledge of noise correlation matrix, which means that it can be used in various applications and is more robust. Furthermore, the proposed algorithm is suitable for practical application due to its low computation complexity. Numerical simulations show that our proposed algorithm for blind separation of correlated sources in noisy mixtures is more effective than existing algorithms.