🚀 go-pugleaf

I have an application that requires me to encode human speech at relatively
low bit rates (<=9600bps) and not require a dedicated DSP processor.  I'm
looking for a method that could work on your typical 33MHZ 486-based PC.  I
realize that this may be a tall order, and certainly do not require that the
speech be telephone quality -- merely intelligible would suffice for starters.

I've looked into PCM, CVSD, and ADPCM, and none of these methods would seem
to work at the bit rates I'm looking for.  I briefly looked at CELP coding,
but that would seem to require more processor bandwidth than I wish to use.
Finally, PELP (Phase Excited Linear Prediction) has been mentioned as a
coding method that may work given my constraints.  A search of the local
technical bookstore has not revealed any texts that discuss this method.  If
anyone can provide details on how it works, comments on its suitability,
references to texts on it, or other suggestions for coding methods that may
meet my requirements, I will be most grateful.

-- 
Bill Dorsey      "Most people mistake law for justice and authority for
dorsey@lila.com   liberty.  You will hear them talk of 'liberty under law,'
ripem public key  and they are content to see it so deep under the law that
on request        it is completely obliterated." -- Herbert Spencer