Howdy Ross,
> > What is this kind of analysis used for, BTW? Automatic
> speech recognition?
> > (I mostly did _speaker_ or voice recognition, so phonemic
> identification
> > wasn't that important to me.)
>
> Well, it's mostly to see if people can safely be lazy with
> research sounds.
I was really trying to ask about what kind of research would be interested
in the sort of phonemic analysis you had described.
> mp3s are a very convienant format for people to use in many
> places. It's
> much easier to distribute mp3s instead of {RAR,ARC,ZIP,bzip2}
> compressed
> audio files. Portable mp3 voice recorders are becoming cheap
> and available.
> It wouldn't be difficult to make a harddisk recorder that
> saved mp3 instead
> of sampling speech with DAT recorders.
Oog. DAT should be way overkill for speech. Doesn't anyone make a mono
mulaw 8kHz codec like the old NeXT's had? That was lossless, and it had a
really low data rate (doing the math: 8 bit * ~8kHz = 64 kbit/s, compared
to DAT's ~1.5 Mbit/s), and it was designed for speech. Could I make a lot
of money by starting a company that built these? (I know lots of devices
use this sort of tech, but most aren't downloadable to computer, which is
what you guys need, and which would probably be useful to most people
recording speech these days anyway.) It wouldn't be a trivial design, but
it shouldn't be a hard one either.
> Aparantly the interest was generated recently when a
> phonetics publication my
> professor reads published an article that detailed a speech
> study in which the
> speech was recorded on a Sony MiniDisc recorder. Aparantly a
> number of people
> doubted the author's results because no one has established
> what mp3 (and
> other similar) algorithms do to speech.
Those people were right to doubt. MiniDisc is a really crude first
approximation to MP3, from what I've read. If sounds imperceptible to
humans are at all important to the analysis, then you (probably) shouldn't
be using psychoacoustic based encoding. Seems like a nice topic for a paper
to demonstrate this...
Hope that helps,
Alex
_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder
