[sdiy] vocoder filters

[Bernard Arthur Hutchins, Jr]

My asking what anyone hoped to accomplish with something dubbed a “Vocoder” was meant to suggest that anyone undertaking such a task (particularly if doing it in analog!) needs to have studied “prior art” and to state a plan going forward.  Presumably the goal is NOT to achieve a communication channel, or to reproduce a (classic – 60 year old) Vocal Encoder/Decoder for a quirky “robot voice”.

I would hope the model was rather an attempt to capture the spectral shape of a segment of speech or singing (most likely of successive “frames” of perhaps 50 msec each) and impose this shape on an entirely new (artificial) excitation.  For example, you might in two second say: ”Dogs and cats each hates the other” and analyze the resulting 40 frames into a sequence of formants - “40 frequency responses”.  If we HAD also analyzed for the original excitation (like pitch - not just the formants), and used this to excite the filter profiles in sequence, we might have some garbled notion of cats and dogs.  Played back with some synthetic substitute excitation (perhaps a saw or noise) we would hope for some vague feeling of a vocal origin capturing our interest if not imparting any verbal information.  But presumably we were not intending to RECOVER speech, so the achievement of a NEW vocal-like impression is what we wanted.

This experiment (and use) is at least 45 years old.  Better name than vocoder?

- Bernie

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20190902/5ee1eaf1/attachment.htm>