Re: Spectral Analysis & Resynthesis of Audio Signals.. Anyone else into this? Applications/Tools that reliably produce good results?

On May 31, 3:30 am, Michel Rouzic <Michel0...@xxxxxxxx> wrote:
dbd wrote:
In what sense of 'resolution' is the reassignment method an
'ultra high resolution' method, as opposed to the many other
STFT based methods?

It's supposed to be a "super-resolution" method (and not "ultra high")
because it's supposed to allow to 'beat' the STFT's (and the likes)
resolution limitations.

That depends on what you consider to be the STFT's
limitations. Since a gapless sequence of complex STFT frames
preserves complete information (above the numerical noise)
about a suitably bandlimited signal represented by the sample
vectors (due to the possibility of reconstruction), it should
have no intrinsic limitation in resolving anything
representable by that signal's information content.

Most of the various methods just look at isolated sections
of the STFT (one bin, three bins), or reweight the signal
information to compress more information into a single STFT
bin of higher degree (offset window weights for phase
vocoder, differently shaped windows weights for reassignment
and derivative methods, etc.), and combine those small subset
results in various ways (linear, parabolic interpolation,
phase difference or magnitude ratio adjustments, etc.)
This allows some bounded computational efficiency or cost
per result.

The primary contribution to accuracy of the estimation
result might be the amount of "relevant" information that
can be compressed into the result. The various methods
mainly just choose different (and perhaps multiple)
weightings of the data which are considered relevant to
that characteristic which one is trying to resolve. The
rest of the information is thrown away. If the thrown
away information was important to the result, resolution
suffers.

..


IMHO. YMMV.
--
rhn A.T nicholson d.0.t C-o-M
http://www.nicholson.com/rhn/dsp.html
.