机械声源信号的带通滤波盲分离

Blind Separation of Mechanical Sounds Based on Band-Pass Filter

盲源分离是一个很独特的盲信号分析与处理工具,在机械设备状态监测与故障诊断领域有较好的应用前景。但在实际应用于机械源信号分离中,效果尚不够理想。特别是,当传感观测中存在高斯噪声或调制源时,往往无法获得准确的源波形恢复。本文在分析现有盲源分离算法的基础上,提出一种新的、基于带通滤波的改进盲源分离方法。仿真以及实际的机械声源信号分离实验结果证实了新方法的有效性。

Blind sources separation is a special tool for analyzing and processing signals blindly, which is promising in condition monitoring and fault diagnosis of machinery. However, the effect of sources separation by such method is not adequate for mechanical sources in practice. Especially, accurate restoral of sources waveform is often impossible when there are Gaussian noise or modulatory sources in sensor measurement. New method to blind sources separation based on band-pass filter has been proposed in this paper on the basis of analyzing the present methods of blind sources separation. Some simulations and the results of separating mechanical acoustic signal from real world verify new methods.