How to go from sound analysis to sound synthesis?

Hi, Everyone!

If I have analyzed a sound and have its spectrogram, how can I utilize
that information for sound synthesis?

Thank You Very Much.

Truly Yours, Simon Dexter

simondex@yahoo.com skrev:
> Hi, Everyone!
>
> If I have analyzed a sound and have its spectrogram, how can I utilize
> that information for sound synthesis?

By computing the average spectrum and compute the Inverse
Fourier Transform?

This is an impossible question to answer, unless you provide
some more information:

- Do you have power spectrograms or magnitude/phase spectrograms?
  Some times you ned the phase of the spectrum.
- Are the data stationary or are they transients?
  Stationary data can to some extent be synthesized from spectra,
  transients can be difficult.
- What is the application you have in mind? The constraints
  imposed by the appliications can completely govern what
  DSP methods can be considered. What can be used in off-line
  processing need not be suitable for on-line processing.

Rune

<simondex@yahoo.com> wrote in message 
news:1124718668.266876.253260@g47g2000cwa.googlegroups.com...
> Hi, Everyone!
>
> If I have analyzed a sound and have its spectrogram, how can I utilize
> that information for sound synthesis?
>
> Thank You Very Much.
>
> Truly Yours, Simon Dexter
>
Simon,

Doing what you want can be very challenging.  Here are some of the reasons:
- you may need to know the analysis bandwidth / resolution that will be 
applied to the synthesized sound.  That's because you need the record to be 
long enough to not end up with a synthesized sound that is made up of 
"lines" while the original was more continuous.
- you need to know what part of the spectrum is really broadband noise so 
that it can be replicated adequately.

Take a look at speech synthesis.  There's a huge body of literature.  The 
objective there was to compress speech and re-create it.  So, there's a 
short-term aspect to the methods - in contrast to the issue I mention above.

Fred