Phase space reconstruction based on stationary wavelet transform

Zhang Lijun,Zhang Lixin,Yang Jianhong,Li Min

DOI: https://doi.org/10.16450/j.cnki.issn.1004-6801.2012.06.011

2012-01-01

Abstract:Aiming at feature extraction of nonlinear signals, a novel phase space reconstruction method based on stationary wavelet transform is proposed. The signal is decomposed into different scales by stationary wavelet transform, and the wavelet coefficients can be used to reconstruct the phase space. The low-dimensional morphological feature which is hidden in the high-dimensional data in phase space is obtained through the Local Tangent Space Alignment algorithm. Experimental results show that, after stationary wavelet transform, the architecture of attractor trajectory is similar to the original one and the low-dimensional morphological feature of nonlinear signal with Gaussian white noise can be extracted by Local Tangent Space Alignment algorithm. And the low-dimensional morphological feature is also similar to the attractor trajectory. The proposed phase space reconstruction method has much more merits than the traditional methods. The proposed method can be used to extract the fault feature of mechanical vibration signals with noise.