基于预白化和子空间方法的主动声呐混响抑制

Active Sonar Reverberation Suppression Based on Pre-whitening and Signal Subspace Method

海洋混响作为主动声呐目标探测时的主要背景噪声,是一种有色干扰噪声,常被看作是非平稳的过程。基于信号子空间分解的噪声抑制算法被广泛应用于语音增强领域,其对于包含加性高斯白噪声的信号有明显的降噪效果,但是对于非高斯分布的混响抑制效果不佳。因此,本文提出首先利用自回归模型构造白化滤波器,对混响进行预白化处理之后再使用子空间分解法,实验数据分析证明,该方法可以明显提升混响的抑制效果。

Reverberation is the dominant background interference. It is a colored disturbance noise and always regarded as a non-stationary random process. The noise suppression method based on signal subspace decomposition was widely used in speech enhancement, and it can effect perfectly on the signal which mainly disturbed by white Gaussian noise, but the effect on underwater acoustics signal which mainly disturbed by reverberations is not so good. In this paper, an autoregressive process of reverberation sequence is modeled to form whitened filters. Then the whitened sequences are processed based on signal subspace decomposition, it can be found that the effect of reverberation suppression is better improved.