An enhanced empirical wavelet transform for noisy and non-stationary signal processing

As an alternative method of empirical mode decomposition (EMD), the empirical Wavelet transform (EWT) method was proposed to realize the signal decomposition by constructing an adaptive filter bank. Though the EWT method has been demonstrated its effectiveness in some applications, it becomes invalid in analyzing some noisy and non-stationary signals due to its improper segmentation in the frequency domain. In this paper, an enhanced empirical wavelet transform method is proposed. This method takes advantage of the waveform in the frequency domain of a signal to eliminate drawbacks of the EWT method in the spectrum segmentation. It modifies the segmentation algorithm by adopting the envelope approach based on the order statistics filter (OSF) and applying criteria to pick out useful peaks. With these measures, the proposed method obtains a perfect segmentation in decomposing noisy and non-stationary signals. Furthermore, simulated and experimental signals are used to verify the effectiveness of the proposed method.