Re: Intelligibility of reversed speech, Why?

[John Hershey <jhershey@xxxxxxxxxxxxxxx>]

It has to do with the information-bearing components of speech.  In machine
perception it has been found that the power spectrum of short windows are
important for speech recognition, whereas the phase is not.

Time reversal of a segment corresponds in the Fourier domain to reversal
(complex conjugate) of the phases of all the components.    Although
windowed time-reversal distorts the phases, it preserves the power spectrum
of the signal within the window.

The best window size for speech perception then should be smaller than the
length of a phoneme, but not so small that the frequency resolution drops
below that of the formants.  However, an important confound in demos of this
type is the noise of discontinuities and aliasing caused by the window which
makes time-reversal using smaller windows harder to listen to.

----- Original Message -----
From: "Yadong Wang" <ydwang@ELE.URI.EDU>
To: <AUDITORY@LISTS.MCGILL.CA>
Sent: Wednesday, January 24, 2001 9:07 AM
Subject: Intelligibility of reversed speech, Why?


> Dear Auditory list,
>
> Saberi & Perrott ("Cognitive restoration of reversed speech," Nature 398:
> 760, 1999)
> recently showed that local time reversal does not destroy intelligibility,
> if the
> time reversal window is brief.
>
> (Please go to http://www.utdallas.edu/~assmann/TREV/trevdemo.html for the
> details.)
>
>
> Why is that?
>
>
>
> Thanks.
>
> -Yadong            Toward a Man-made Brain
> ___________________________________________
> 401-874-5392 (O)       Leave Msg Please...
> 401-789-7742 (H)
>
> Dept of Electrical  & Computer Engineering
> HTTP HomePage         http://www.CoZeC.com
>