In order to reduce the noise

a denoising method of DT-CWT（Dual-tree complex wavelet transform） is proposed to extract the weak failure information in machinery vibration signals to diagnose mechanical fault.The dual-tree complex wavelet transform filter design principles are described.The presented method can not only avoid gibbs phenomen

but also obtain higher signal-to-noise ratio（SNR） than the common methods based on orthonormal discrete wavelet transform

because of improving asymptotic optimality of shift invariance of DT-CWT.The experiments indicate that this method can extract the fault feature submerged in a heavy noise.