Return to the homepage
Search the complete Wavelet Digest database
Help about the Wavelet Digest mailing list
About the Wavelet Digest
The Wavelet Digest
> Volume 5, Issue 6
Question: Implementation of Mallat algorithm

Previous :: Next

Author 
Message 
Andy Woodward (azw@aber.ac.uk) Guest

Posted: Fri Aug 09, 1996 1:37 pm Subject: Question: Implementation of Mallat algorithm




Question: Implementation of Mallat algorithm
Something has been confusing me about some implementations of the
Mallat algorithm.
The big advantage of the Wavelet Transform over the FFT is its ability
to deal with nonstationary signals. In some versions of the Mallat
algorithm, the filter ladder is implemented in the frequency domain
relying on the Convolution Theorem (that a convolution in the time
domain, as carried out by a Toeplitzlike matrix, is equivalent to a
multiplication in the frequency domain). BUT in order to transform to
the frequency domain to carry out this procedure, an assumption must
be made that the data at that scale is stationary or one of the
fundamental assumptions of the FFT (and hence of the Convolution
Theorem) is violated. So it seems to me that the wavelet pcoefficients
calculated in this way must reject the nonstationarity that is one of
the main reasons for using the WT in the first place?
I assume I am misunderstanding something fundamental here  what????
Is it valid to calculate the Mallat algorithm in the frequency
domain, or only (as it seems to me) exclusively in the time domain? 





All times are GMT + 1 Hour

Page 1 of 1 
