基于非负矩阵分解的欠定卷积盲源分离方法

摘要

Aiming at underdetermined convolutive blind source separation,a method based on nonnegative matrix factorization (NMF) is provided.The STFT of each source signal is given by gaussian components,with the maxim of Itakura-Saito divergence as target function.The mutiplicative update (MU) is used to estimate the original signal in the frequency domain for improving the accuracy.The simulation result verifies the efficiency of the method.%针对音频信号欠定卷积混合模型的盲源分离求解问题,提出一种基于非负矩阵分解(NMF)的盲源分离方法.该方法以板仓-斋藤(Itakura-Saito)散度和的最大值为目标函数,利用高斯分量表示源信号的短时傅里叶变换(STFT),使用乘积更新算法估计频域内的源信号,以提高其估计的准确度.仿真结果验证了该方法的有效性.