中文摘要：

频域盲源分离方法通过STFT变换将时域的线性卷积模型转化为频域的瞬时混合模型，可以利用瞬时混合的成熟算法，然而缺点是存在排列和尺度的不确定性，会使逆STFT变换后的信号发生再次混叠。对分离矩阵内部结构进行研究，提出邻频幅角比的概念，通过纠正图样中发生排列错误频点处的分离矩阵结构，达到正确拟合已分离信号的目的。仿真结果表明，邻频幅角比排序算法可以纠正大多数频点上的排序错误，正确进行盲源分离。

英文摘要：

The Blind Source Scparation（BSS） can turn linearity convolution model in time domain into the instantaneous fixed one in frequency domain by STFT transform.The traditional instantaneous fixed algorithm has indeterminacy in permutation problem and scaling problem,so as to make the signal mixed again after the converse STFT transform.In this paper,based on the study on the structure of separated matrix,a conception of neighbor frequency breadth-angle ratio is proposed,which may contribute in correcting the structure of separated matrix in the frequency that occurs permutation errors.The simulation results show that such compositor algorithm based on neighbor frequency breadth-angle ratio can correct most permutation errors and conclude an exact signal in the blind source separation.