Frank, agreed, I would never criticize an honest question, but I might challenge an ill-founded assertion.
Separate issue -- The HelloDirect reference reminded me of a question I have pondered for some time, but never answered. Someone on the thread may have some insight.
As a hypothetical situation, many separate "HelloDirects" are out there providing compressed voice service to customers from various gateways, each offering some flavor of cheap access. I further want to assume this happens before (or flatly without) any industry consensus on compression schemes/rates. So, one company is using 11kb vocoders of some kind; another is using 11kb of another kind, a third is using 6kb vocoders with the same type compression as the first, and so on. In this environment, it seems like there is either a hierarchy of conversions, or more likely in this business environment, this is more of a mesh (mess?). The question is: what happens to voice quality that is perfectly acceptable with two voice conversions of like type, the PCS situation, when there are back-to-back conversions of different type compression? In particular, what happens to code excited linear predictive voice coding when the input is decoded speech from a different vocoding system? Are there relevant experiments that have been conducted and published? Am I just imagining a problem that is no problem at all? I just have an instinct that when the limits are pushed, with some of the really aggressive vocoding that is possible today, that it might collapse if we mix and match to make a complete connection through multiple concatenated systems. Thread?
|
