FILTER BANK SYNTHESIS FOR ADAPTED WAVELET ANALYSIS
Zoltán Germán-Salló · 2010
The multiresolution theory of orthogonal wavelets proves that any conjugate mirror filter characterizes a wavelet that generates an orthonormal basis of L 2 R. This paper presents a method to obtain a discrete wavelet function, through FIR filter synthesis starting from a random discrete sequence. This sequence is used as a FIR filter which generates the four conjugate mirror filters necessary to implement the Discrete Wavelet Transform. The refinement of this sequence is carried out by a learning structure which follows to minimize the error between the signal and its reconstructed form from first order decomposition.