On the energy leakage of discrete wavelet transform

The energy leakage is an inherent deficiency of discrete wavelet transform (DWT) which is often ignored by researchers and practitioners. In this paper, a systematic investigation into the energy leakage is reported. The DWT is briefly introduced first, and then the energy leakage phenomenon is described using a numerical example as an illustration and its effect on the DWT results is discussed. Focusing on the Daubechies wavelet functions, the band overlap between the quadrature mirror analysis filters was studied and the results reveal that there is an unavoidable tradeoff between the band overlap degree and the time resolution for the DWT. The dependency of the energy leakage to the wavelet function order was studied by using a criterion defined to evaluate the severity of the energy leakage. In addition, a method based on resampling technique was proposed to relieve the effects of the energy leakage. The effectiveness of the proposed method has been validated by numerical simulation study and experimental study.