[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Intelligibility of reversed speech, Why?
Are you asking how does intelligibility (measured as percent correct)
vary as a function of the duration T of the window of reversal? Surely
that data must exist since someone has written here that intelligibility
seems to hold up for T up to the length of a phoneme, say, 50 ms? It
would be interesting to see also for several durations T (say 10, 30,
50, 100 and 200 ms) how intelligibility varies as a function of signal
to noise ratio. Pick your noise: white, pink, speech shaped, babble.
Acoustic information in speech is redundant, of course, but certainly,
as information (a cue of what ever sort) was removed or otherwise
blurred, you would expect the intelligibility to decrease for a given
S/N ratio. Removing cues would then be essentially the same as adding
more noise. Wouldn't it?
----- Original Message -----
From: Tóth László <tothl@INF.U-SZEGED.HU>
Sent: Thursday, January 25, 2001 9:57 AM
Subject: Re: Intelligibility of reversed speech, Why?
> On Wed, 24 Jan 2001, Yadong Wang wrote:
> > local time reversal does not destroy intelligibility,
> Can somebody tell me how exactly the statement "transformation T does
> destroy the intelligibility of speech" is defined?? I think one should
> very careful with such a statement. My idea is that speech cues are
> redundant. This means that after the removal of one cue the speech
> may retain its intelligibility - under the given cirsumstance (e.g.
> environment). This does not necessarily mean that the removed cue
> not help intelligibility under different circumstances (e.g. in
> So I think that the statement "transformation T does not destroy
> the intelligibility of speech" should be defined as "sentence X and
> sentence T[X] has the same intelligibility UNDER ANY POSSIBLE
> CIRCUMSTANCES". (Let's measure "same intelligibility" using some
> well-defined psychoacoustic experiment). To put it an inverted way,
> "transformation T DOES dicrease intelligibility if there is at least
> experimental setting in which the intelligibility of the transformed
> signal is decreased compared to the original one."
> I would apply this definition to the good old statement about the
> deafness of the ear. Was it ever exhaustively and thoroughly examined
> phase information under no circumstances can serve as a cue and
> Any opinions? (and sorry for thinking as a mathematician, but I can't
> it any other way :-) )
> Laszlo Toth
> Hungarian Academy of Sciences *
> Research Group on Artificial Intelligence * "Failure only begins
> e-mail: email@example.com * when you stop
> http://www.inf.u-szeged.hu/~tothl *