摘要
针对欠定盲源分离问题,从理论上分析矩阵对角化方法存在的不足。在混合矩阵已知的基础上,提出时频邻域源个数精确估计和矩阵对角化方法的源信号估计算法。通过度量不同的时频邻域内源信号估计结果的协方差矩阵的对角化程度,同时估计当前多源邻域内源信号数目及其时频波形。最后通过对混合矩阵求伪逆完成源信号的估计。理论分析和仿真结果表明该算法能够提高源信号的估计性能,并提高对源信号数目的适应能力。
Aiming at the problem of underdetermined blind signal separation,the drawback of the covariance matrix diagonalization is analyzed,and a method of identifying the number of time frequency neighborhood source and an algorithm of separating source signals are proposed based on matrix diagonalization.By calculating the differences between the diagonal elements and the non-diagonal elements of the covariance matrix of the estimated source signals,the real number of active signals is estimated in each multi-source neighborhood,as well as their corresponding mixing matrix in different time frequency neighborhoods.Finally,all the sources are separated by calculating pseudo-inverse of the mixing matrix.Theoretic analysis and simulation results indicate that the source signals with higher gain are separated with the proposed method compared to the other algorithms.
作者
王翔
赵雨睿
李保国
WANG Xiang;ZHAO Yurui;LI Baoguo(College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China)
出处
《电子信息对抗技术》
北大核心
2021年第1期1-5,32,共6页
Electronic Information Warfare Technology
关键词
欠定盲分离
短时傅立叶变换
通信对抗
侦察
矩阵对角化
源个数估计
underdetermined blind source separation
short time Fourier transform
communication counter
reconnaissance
matrix diagonalization
source number estimation
