摘要
针对传统盲分离算法在以欠定的观测信号为对象时的失效问题,提出一种改进的结合四阶累积量的二阶联合对角化盲分离算法.该算法通过对源观测信号进行四阶累积量计算,比较各累积量切片对算法分解的合理性,对形成的多个切片进行近似对角化处理,进而对以延迟点为中心进行多个对角化矩阵的平均,提高算法分解的稳定性,最后通过仿真信号和车辆振动信号验证算法的有效性.该联合算法有效扩展了二阶音分离(SOBI)算法的应用范围,可以应用于瞬时混合模型的实际工程信号处理.
Based on the inability of the traditional blind source separation method on the analysis of the undermined observation signal,an improved second-order joint diagonal based FOC（four order cumulant）algorithm was proposed.Firstly,the observed signal was calculated using the four order cumulant,and the diagonalizable rationality of the different cumulant slice matrix was compared according to the second-order algorithm.Followed by the further treatment of the different cumulant slice matrix and average of the diagonalized matrixes on multiple time-delay points,the numerical stability of new algorithm was improved.The effectiveness of the method combining the improved second-order algorithm and average of matrixes was verified by the simulated signal and vehicle vibration signal.The combined method effectively expands the application scope of SOBI（second-order blind identification）,which can be further applied to the actual engineering signal of the instantaneous mixture models.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2016年第7期86-90,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51405221)
江苏省自然科学基金资助项目(BK20130746)
南京工程学院校级科研基金资助项目(YKJ201334)
关键词
盲源分离
四阶累积量
联合对角化
切片
延迟点
二阶音分离
blind source separation
four order cumulant
joint diagonalization
slice
time-delayed point
SOBI（second-order blind identification）
