limitations of MD

[Celso Alvarez Caccamo]

P L Patrick wrote:
>         ... but it is my understanding that MD technology is not
> useful for anyone who plans to do acoustic analysis of
> sound recordings later
> on, or to preserve that possibility for other researchers, due to the
> digital compression that alters the signal beyond the possibility of
> restoration to the original spectral characteristics.

I'm not sure about that. Only extremely low and high frequencies are
trimmed. MD frequency range approaches that of a CD, which is
20 - 20.000 Hz, much wider than the spectrum of human speech.
In fact, I understand that spectrographic analysis works better if
you resample audio to 8.000 or 11.500 Hz.

My impression is that MD quality is more than enough for
speech analysis. It is true that DAT is better, though, but
I doubt that analog is better, as it introduces noise distorsion
which is hard to clean up. With MD, at least, you have cleaner
recordings, which you can easily digitalize directly as WAV
through the computer's Line In, with little noise.

MP3 compression also operates on the 20 - 20.000 Hz range, as
do digital formats such as WAV or the CD audio tracks. MP3
players such as Yepp also record voice in compressed format,
but the quality is still very bad for any type of linguistic
work -- actually, recording is not MP3, but some other type.

Yes, I think that the future of digital recording is MP3,
not as much in terms of quality (DAT is better) as in terms
of direct digital transfer to computer. Or, else,
MD will finally open up to the computer industry
and distribute software for direct extraction and conversion
of MD tracks to WAV files. I don't know why they are not
doing it yet.

Uncompressed-sound recorders such as CD tracks or WAV are
likely not the way to go, I believe, because of the very large
files they produce -- at least until the storage media is
cheaper. Even though the technology is already here
for large, solid-state storage media, commercial interests
(the CD-manufacturing industry, for example) prevent its
development. 64 MB of a MP3 player, for example, can hold
upt to 1 hour of compressed stereo sound, but only 1/10th in
uncompressed WAV. So, I believe it is more economical still
to record compressed sound and uncompress it only for
speech analysis and other purposes. To my knowledge, there
is still no audio software that works on compressed sound
directly, not even compressed WAV. They all (CoolEdit, for
example) perform some sort of uncompression.

Another problem for recording uncompressed WAV (that is,
the format for CD-recorders/rewriters) is that, even though
in speech analysis phoneticians work with mono files, the
format of CD tracks (file extension .CDA, but identical
to WAV) must always be 44.100 Hz, stereo. So, you don't save
space by recording mono, as the file must have two tracks,
even if identical.

Finally, another compression format which I like a lot is
RealAudio. In its latest release, RealEncoder can compress
at high frequency resolutions, even higher than MP3, I
believe. There are a number of freeware utilities that
convert RealAudio to WAV. RealEncoder can encode and compress
in real time through the Line In if your computer is fast
enough. RealAudio files are "streamable" through the Internet,
and they are slightly smaller than MP3. Other utilities
allow you to cut and paste RealAudio files. With RealAudio,
by using mono instead of stereo you do save disk space.
Real is at http://www.real.com .

-celso

--
Celso Alvarez Cáccamo              Tel. +34 981 167000 ext. 1888
Linguística Geral, Faculdade de Filologia     FAX +34 981 167151
Universidade da Corunha                          lxalvarz at udc.es
15071 A Corunha, Galiza (Espanha)  http://www.udc.es/dep/lx/cac/