[PD-ot] sine wave speech

Bryan Jurish moocow at ling.uni-potsdam.de
Tue Sep 18 09:38:54 CEST 2007


moin all,

On 2007-09-17 23:17:42, Ed Kelly <morph_2016 at yahoo.co.uk> appears to 
have written:
> In theory there are 5 formants in speech, spaced at roughly 1KHz 
> intervals: 60-1000Hz, 1000-2000Hz etc (roughly).
> If you make sine waves in these frequency ranges you can make vowel-like 
> sounds.

well, 6th and even 7th formants have been (pretty convincingly) argued 
for; but of course they get less prominent the higher you go: you can 
make pretty intelligible vocalic sounds with just 3 formants...

> I don't know what they would do for consonants though - FM?

you could use a noise source (e.g. for fricatives) in parallel with the 
vowel generator and a mess of band-pass filters.  See:

   D. H. Klatt (1980) "Software for a cascade/parallel formant
   synthesizer.", Journal of the Acoustical Society of America 67(3),
   March 1980, pp 971--995.

... which is (I believe) available in scanned form on the net somewhere.

> Haven't read the paper - this is a glib answer! If and when I get this 
> formant analyser I want to build made and working, you should be able to 
> drive sines with it, to talk in sines. But if you use Snack tools or 
> something to get formant analyses of speech, you should be able to load 
> them into a msgfile and drive sines from that.
> 
> ...but I need to get a lot finished before I can turn to that again. How 
> to get a resonance curve from filter coefficients in 1 block is my 
> problem...
> 
> Ed

Let us know when you get things running!

marmosets,
	Bryan

> 
> */IOhannes m zmoelnig <zmoelnig at iem.at>/* wrote:
> 
>     Hans-Christoph Steiner wrote:
>      > Hey all,
>      >
>      > I just found this paper, and I wonder if anyone has implemented a
>      > sine-wave speech synth in Pd.
>      >
>      > http://www.scholarpedia.org/article/Sine-wave_Speech
>      >
>      > Basically, you can make intelligable speech using just three sine
>      > waves.
> 
>     i only remember having used something as simple (and obvious) as
>     [fiddle~] with 5 partials or so to do this as one of my very first test
>     with this object (back in 1998 or so)
> 
> 
>     fmasdr.
>     IOhannes

-- 
Bryan Jurish                           "There is *always* one more bug."
jurish at ling.uni-potsdam.de      -Lubarsky's Law of Cybernetic Entomology




More information about the PD-ot mailing list