[Lexicog] Re: Audio device for field language data collection

[Joan Baart]

Hi all,

I know of several studies that address the concern that audio compression
algorithms may distort the audio signal in a way that makes it less useable
for acoustic analysis of speech. An example of such a study is Van Son
(2002). Van Son observed in his experiments (looking at variables that are
typically of interest to phoneticians) that the distortion caused by
compression and subsequent decompression of speech (he looked at MP3, Ogg
Vorbis and Minidisc) is not greater (and generally much smaller) than that
caused by a change of microphone. Choice of (good-quality) microphone varies
widely between projects and is usually not seen as a problem. Van Son's
conclusion is, therefore, that decompressed speech can be considered
acceptable for most acoustic phonetic applications. See:

http://www.fon.hum.uva.nl/rob/Publications/CoCOSDA2002paper.pdf

and also:

http://www3.uji.es/~gonzalez/J%20Voice.pdf

It has been two years since I needed to look into this subject. If anybody
is aware of more recent work that perhaps changes the picture, then I would
be interested.

Joan

----- Original Message ----- 
From: Victor E. Roetman
Sent: Monday, September 13, 2004 4:03 AM
Subject: [LTS]Re: compressed audio for acoustic phonetics (WAS: Audio device
for field language data collection)


I'm no expert on this, either, but when I was in NDSIL 2002, Peter
Ladefoged and others were talking about some of these things, and they
all seemed to really like minidisk, and they seemed to have no problems
with measuring anything related to speech.  Minidisks use ATRAC
compression, which is similar to mp3, so I expect mp3 would not be any
different.  Plus, you can choose bit rates of greater than 128kb/s,
variable bit rate compression, etc, and get better results.  There is
also things like FLAC lossless compression for .wav files which
typically does somewhere around 50% compression.

Check out http://www.avisoft-saslab.com/compression/compression.htm for
some information, and regarding lossless compression methods you could
look at http://flac.sourceforge.net/comparison.html

vic


---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.756 / Virus Database: 506 - Release Date: 9/8/2004
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lexicography/attachments/20040913/93deca4a/attachment.htm>