Re: Speech(or Phase) Reconstruction from Magnitude Spectrum

[Toth Laszlo <tothl@xxxxxxxxxxxxxxx>]

On Fri, 28 Jan 2005, ismail uysal wrote:

> Hence my question is...Is it possible to reconstruct speech using only
> magnitude spectrum? Any comments will be greatly appreciated.

You didn't mention whether you have a narrowband or wideband spectrum.
With a narrowband spectrum you can try to track the harmonics and take
care to keep the phase coherent from frame to frame. With this method you
can obtain guite a good (although not perfect) reconstruction. For more
details search for the sinusoidal (or "harmonics plus noise") model of
McAulay and Quatieri (or Serra and Smith).

               Laszlo Toth
        Hungarian Academy of Sciences         *
  Research Group on Artificial Intelligence   *   "Failure only begins
     e-mail: tothl@xxxxxxxxxxxxxxx            *    when you stop trying"
     http://www.inf.u-szeged.hu/~tothl        *