Date   : Sat, 11 Mar 2006 17:45:24 +0000 (GMT)
From   : Sprow <info@...>
Subject: Re: 101 uses for a speech system (number 35)

In article <4412DABE.3070404@...>,
   Fragula <fragula@...> wrote:
> Sprow wrote:

> > which describe the US dialect speech ROM
> Well good heavens.. i'd more or less forgotten about this bit of the beeb!
>
> Oooh! Anybody ever hear the fabled "Alan Alda" (from M.A.S.H.) PHROM?
>
> There were four different U.S. ones. I have word listings for them
> someplace.

That'd be interesting to get hold of.

> > A little known fact is that the speech PHROM which normally holds speech
> > data can also be used to hold RFS data, 
>
> heh. I knew that. Wasn't actually that usefull a bit of info.

Not at the time, since the original PHROMs were mask programmed parts so you
had to buy many 1000's at once: rather impractical.

> Does anyone know much about the internal format of the (original) 6100?
> I mean as a speech rom, rather than RFS. I know it's something like
> "waveform type, frequency and a bunch of (vocal tract) coefficients" per
>  "chunk" (which could be sound primitives, or whole words, or a bunch of
> "chained" primitives to make synthetic words. It was kinda complex, that
> I do recall.

It's covered in rather dry terms in the Acorn User Guide and TMS5220
datasheets (both available to download), and basically uses LPC-10
compression but with some Texas tweaks such as replacing mute time with a
special zero energy marker and repeated runs for simple run length encoding.

> Ahh.. I think I remember how. TI extended BASCI had a function to "read
> PHROM" into strings, but used "logical" addresses (i.e. phrase numbers)
> or somesuch.

There's a similar SOUND command to access words based on word numbers,
Sprow.