[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Intelligibility of reversed speech, Why?

On Thu, 25 Jan 2001, Jont Allen wrote:

> It is clear there is redundancy in once sense. Language is an entropy
> constrained code.
> One should not be testing with high-context low-entropy sentences. One
> should be using nonsense words, or other ways to remove such context
> effects.

When I said "redundancy of speech", I was especially thinking of
redundancy in the acoustic coding of the phonemes, and not linguistic

> When one uses high entropy words, then I really wonder if speech is redundant.
> My feeling is that it is not.
I'm trying to find some convincing example, but it is not easy. The best I
can find is the recognition of plosives. It is said that there are (at
least) two cues that help us classify plosives: one is the shape of the
burst itself (the place of the "locus"), another one is the transition of
the formants (especially the 2nd) in the following vowel. As I remember,
there were tests reporting that if the bursts themselves are replaced with
white noise, speech remains intelligible (I don't know if these tests used
nonsense words). However, the shape of the burst must be an important cue
in itself, since plosive are not always followed by wovels, so we can't
always look for the formant transitions. Isn't this redundancy?

> One problem with redundancy is that what do you do when the two redundant
> cues give you different messages?
You can resolve the redundancy somehow - worst case by totally missing
the word. My example would be the McGurk effect.
Artificially one can create such examples that have contradicting cues,
and check what happens. Teh problem is, if you don't exactly know what the
speech cues are, you cannot tell if misrecognition was caused by
contradiction of cues or because of missing cues.

               Laszlo Toth
        Hungarian Academy of Sciences         *  "In our life there's if
  Research Group on Artificial Intelligence   *   In our beliefs there's lie
     e-mail: tothl@inf.u-szeged.hu            *   In our business there's sin
     http://www.inf.u-szeged.hu/~tothl        *   In our bodies there's die"