基于Gabor变换的特征提取方法

Feature Extraction Based on Gabor Transform

鉴于旋转机械故障特征信号具有周期性或近似周期性,提出了基于Gabor变换的特征提取方法。该方法首先利用信号构造矩阵,再根据Gabor展开的线性时频变换特性,均值化Gabor展开系数,然后利用Gabor逆变换重构信号以实现降噪。仿真和试验结果表明,该方法能有效地提取故障信息,在旋转机械故障诊断领域应用前景广泛。

In view of the periodic or approximately periodic characteristics in the failure signal of rotating machinery, a feature extraction method is presented based on Gabor transform. At first, the Gabor transform is applied to the structure matrix , and the matrix is composed of the signal. According to the time - frequency linear properties in Gabor transform domain, the Gabor expansion coefficients are processed by the average in the groups of Gabor coefficients, then the de - noised signal could be obtained by the inverse Gabor transform.. The simulation demonstrates that the averaging method based on Gabor transform is efective in signal processing, and could provides good prospects of application in failure analysis of rota- ting machinery.