基于最大信噪比的盲源分离算法

提出一种新的低计算复杂度的瞬时线性混叠信号的盲分离算法,该算法利用统计独立信号完全分离时信噪比量大作为分离准则。源信号用估计信号的滑动平均代替,把源信号和噪声信号协方差矩阵的函数表示成广义特征值问题,通过广义特征值问题求解分离矩阵不需要任何迭代运算。和典型的信息理论方法相比,该算法的优点是具有非常低的计算复杂度。计算机模拟实验证明,该算法能够分离线性混合的超高斯和亚高斯源信号,并且可以有效地分离语音信号。

A Blind Source Separation Algorithm Based on Maximum Signal Noise Ratio

A novel low computational complexity instantaneous linear mixture blind separation algorithm is proposed, which is based on the character that Signal Noise Ratio (SNR) is maximal when statistically independent source signals are completely separated, and it is used as a separation contrast. Source signals are replaced by moving average of estimate signals. The function of covariance matrixes of the source signals and noises is expressed by the generalized Eigenvalue (GE) problem, and unmixing matrix is achieved by solving the generalized Eigenvalue problem without any iterative. Compared to the typical information - theoretical approaches, the merit of this new algorithm is very low computational complexity. Computer simulation demonstrates that this algorithm can separate the linear combinations of sub - Gaussian and super - Gaussian sources, and can separate speech signals effectively.