[
http://opencast.jira.com/browse/MH-7600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=26229#comment-26229
]
Stephen Marquard commented on MH-7600:
--------------------------------------
Email thread:
>>> "Stephen Marquard" <[email protected]> 4/15/2011 4:06 PM >>>
Here's some quick data. For an approx 1 hour audio recording (of
speech):
Uncompressed:
WAV: 623M
FLAC: 199M
M4A: 120M (max quality setting in Audacity)
MP2: 71M
MP3: 57M
FLAC did not show any compression gains using zip, MP3 compressed 3%,
WAV compressed 33% to 419M (a lot worse than FLAC).
So in our configuration I'd certainly be prepared to pay the extra size
penalty (71M > 199M) to know that compression artefacts weren't
negatively impacting speech recognition, for example.
Is there a JIRA for this, or should I create one?
Regards
Stephen
--
Stephen Marquard, Learning Technologies Co-ordinator
Centre for Educational Technology, University of Cape Town
http://www.cet.uct.ac.za
Email / IM (Jabber/XMPP): [email protected]
Phone: +27-21-650-5037 Cell: +27-83-500-5290
>>> Christopher Brooks <[email protected]> 4/14/2011 10:58 PM >>>
Hi Stephen,
I understand, I just think the bandwidth will be killer. But
regardless, we did have arbitrary encoding profiles on the roadmap,
though it is now delayed so I'm not sure when it will be coming (we
will be reevluating). This would let you do aac, mp3, flac, etc.;
whatever gstreamer can do (which is a fair bit) and perhaps more.
Regards,
Chris
> AAC is also lossy though. The point of using a lossless
compression
H> in
> the first (capture) phase is not to close-down or complicate
> processing options that work best on uncompress formats.
>
> For example my simple tests on noise reduction with sox show that it
> works very well on WAV audio, but is terrible on mp2 files.
>
> Similarly, speech recognition applications need 16 KHz WAV, and it
> would be better to resample to that from another lossles format than
> to go raw data > lossy format > WAV.
>
> Regards
> Stephen
>
>
--
Christopher Brooks, BSc, MSc
ARIES Laboratory, University of Saskatchewan
Web: http://www.cs.usask.ca/~cab938
Phone: 1.306.966.1442
Mail: Advanced Research in Intelligent Educational Systems Laboratory
Department of Computer Science
University of Saskatchewan
176 Thorvaldson Building
110 Science Place
Saskatoon, SK
S7N 5C9
> Support lossless audio codec for audio capture
> ----------------------------------------------
>
> Key: MH-7600
> URL: http://opencast.jira.com/browse/MH-7600
> Project: Matterhorn Project
> Issue Type: Community Feature Request
> Components: Capture (Devices and Software)
> Affects Versions: 1.1
> Reporter: Stephen Marquard
>
> Capture Agents should have the ability to save captured audio in a lossless
> format, e.g. FLAC, in preference to the lossy mp2 format.
> Lossless audio may be important for downstream applications such as noise
> removeal and speech recognition, which may be negatively impacted by lossy
> formats.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
