卷积混合机械非平稳振动信号的二阶盲分离方法

BLIND SEPARATION METHOD BASED ON SECOND-ORDER STATISTICAL FOR CONVOLUTIVE MIXTURES OF NON-STATIONARY MECHANICAL VIBRATION SIGNAL

针对机械振动信号具有非平稳和卷积混合的特性,文中将基于二阶统计量的盲源分离方法推广至卷积混合的模型,提出在信号子空间的频域中对机械振动信号的盲解卷积方法。仿真和实测数据实验结果表明,此方法充分考虑信号的非平稳以及卷积混合特性,能较好地实现机械振动信号的盲分离。与传统盲源分离算法比较,该方法更适合于机械振动信号的分析。

Mechanical vibration signals in practice can be viewed as stuns of differently convolved source, and the signals are non- stationary. For this characteristic, the blind source separation based on the second-order statistics been expended to the convolution mixture model, a blind deconvolution technique using frequency distribution of signal sttbspace to mechanical vibration signal, is proposed, The experimental result of the numerical simulation data and the actually measuring data shows that the method is efficiency. Compared to the traditional method, this blind de-convolution algorithm adapts to mechanical vibration signal preferably.