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

Re: speech from noise


There is an article which would be useful for you to take a look at if
you are planning on creating speech-like signals with some
intelligibility from noise signals. Actually, in the following article,
we can not create speech-like signal only from the noise. A very small
amount of information from the speech (temporal envelopes of
band-limited speech) should be used to create those kind of signals.

R. V. Shannon, F. Zeng, V Kamath, J. Wygonski and M. Ekelid,
``Speech recognition with primarily temporal cues,''
Science, vol.270, pp.303--304, 1995.

In the article, he conducted speech recognition experiments (not by ASR
but human). Those experiments involved the stimuli which are the noise
modulated with temporal envelopes of speech corresponding to each
frequency band. As a result, more than 90% correct identification of
words, vowels and consonants was observed under conditions of greatly
reduced spectral information.

Hope this reference would be able to help you in some way,
Good luck :-)

 NTT Communication Science Laboratories
 Speech open laboratory
 e-mail: kinoshita@cslab.kecl.ntt.co.jp

hadi harb wrote:

> Hi,
> I've heard that we can create speech-like signals with some
> intelligibility from noise signals. is it true or just I'm wrong?
> if such experiments of generating speech-like signals from other signals
> exist it would be helpful for my research to have references or
> directions toward such experiments
> Thank you,
> Hadi Harb
>  Dept. Mathematiques - Informatique,
>  36, av Guy de Collongue,
>  69130 Ecully France EUROPE
>  tel: +33 (0)
>  fax: +33 (0)
>  e-mail: hadi.harb@ec-lyon.fr <mailto:hadi.harb@ec-lyon.fr>
>  web: http://perso.ec-lyon.fr/hadi.harb