A NEW METHOD OF ESTIMATING INFINITELY SUPPORTED WAVELET FROM A SIGNAL

摘要

This paper describes a novel method of estimating infinitely supported wavelet matched to a signal. The method is based on maximizing the projection of a given signal onto the scaling subspace. This results in energy minimization of the signal in the wavelet subspace so that it can be considered as additional/finer information in the wavelet subspace. Based on this approach, first, the analysis wavelet filter is extracted from given signal. Next, the dual scaling filter is computed by considering orthogonal scaling and wavelet subspace. Corresponding to these FIR analysis filters, synthesis filters are computed. But to satisfy the property of Perfect Reconstruction, resulting synthesis filters are IIR filters that are implemented as anticausal but stable filters. This result in infinitely supported wavelet matched to a given signal.