一种基于累积量的盲源分离新方法及其应用  

A New Blind Source Separation Algorithm Based on Cumulants and Its Applications

作　　者：雷衍斌[1] 李舜酩[1] 郝青青[1] 

机构地区：[1]南京航空航天大学,南京210016

出　　处：《中国机械工程》2010年第7期787-792,共6页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(50675099)

摘　　要：结合二阶累积量和四阶累积量各自的优点,提出一种基于联合近似对角化二阶累积量和四阶累积量的盲源分离算法。采用稳健白化算法有效地减小了噪声对分离精度的影响。盲源分离算法与基于四阶累积量和二阶累积量的算法相比,具有收敛速度快、分离精度高的优点,两个仿真试验验证了该算法能有效分离语音信号和超高斯与亚高斯信号混合的信号。应用该算法成功实现了实测转子复杂混叠振动信号的分离。Combing the advantages of second- and fourth-order, a new algorithm based on sec- ond- and fourth-order cumulants simultaneous diagonallization was proposed, which was based on joint approximate diagonallization. A robust whitening method was introduced, which can effectively reduce the influence on the separation precision. Compared this algorithm＇s performance to JADE al- gorithm＇s and second--order algorithm＇s, the results show that the algorithm has the advantages of fast convergence speed and high separation precision. Through two simulation experiments, the results show that the proposed algorithm can separate speech signals and extract the independent sources from the hybrid mixture of any super--Gaussian and sub--Gaussian signals effectively. At last, this algorithm was applied to separate the practical rotor＇s vibration signals, the rotors＇ complex vibration signals are separated successfully. It provides a new method for separation of complex vibration signals in fault diagnosis processes.

关 键 词：盲源分离 累积量 同时对角化 转子 振动信号 

分 类 号：TN911.3[电子电信—通信与信息系统] TH113[电子电信—信息与通信工程]