匹配追踪法可将信号分解成一系列时频原子的线性组合,并在每次分解前自适应地初始化时频原子的参数。本文研究了匹配追踪的原理和方法,并将匹配追踪后的分解结果与Wigner分布相结合,得到了具有高分辨率的自适应时频分布。通过仿真算例将该法与Wigner分布、小波包变换,短时傅立叶变换的结果相比较,验证了该方法的优越性。最后,将该方法应用于轴承的故障诊断,结果表明,本文方法用于故障诊断的特征提取是有效的。
Matching pursuit decomposes signal into a linear combination of time frequency atoms and adaptively initialize their parameters before each decomposition. We study the principle and method of matching pursuit and obtain the adaptive time-frequency distribution with fine resolution through combining the decomposition results after matching pursuit with the Wigner distribution. Comparative simulation instances show that our matching pursuit method produces better results compared with the Wigner distribution, the short-time Fourier transform and wavelet packet transform. The application of our method to the fault diagnosis of bearings indicates that it is effective for feature extraction during fault diagnosis.