Revolutionary audio compression
John Reidar Mathiassen
If I were to invent an audio compression algorithm that could compress audio
to half the size (or a quarter the size) of e.g. mp3 or ogg-vorbis with the
same psycho-acoustic quality, how useful would it be? Assume that the
computational cost of decoding/encoding is 2-4 times that of mp3 or
ogg-vorbis. Are there any uses for high-compression and "slowly"
encoded/decoded audio compression algorithms?
If a new algorithm for audio compression would emerge, would it not take a
long time before people replaced their mp3's with the new format?
Considering the size of current hard-drives and the steadily increasing
internet bandwidth, how good (in terms of compression-rate, quality
(subjective/objective), decoding/encoding speed) would a new audio
compression algorithm have to be to be considered useful for the masses. Are
there any niche-markets in which such an algorithm would be useful?
Are there any related links/articles in which the mentioned questions have
been discussed?
Regards,
John Reidar Mathiassen
Geoff T
> Hello,
>
> If I were to invent an audio compression algorithm that could compress audio
> to half the size (or a quarter the size) of e.g. mp3 or ogg-vorbis with the
> same psycho-acoustic quality, how useful would it be? Assume that the
> computational cost of decoding/encoding is 2-4 times that of mp3 or
> ogg-vorbis. Are there any uses for high-compression and "slowly"
> encoded/decoded audio compression algorithms?
The only application I can think of immediately is streaming over
low-bandwidth connections. In terms of storage, the cost of harddrives
is such that I've even considered lossless audio compression to store
my albums.
For people who wouldnt mind the longer encode time, quality is usually
the issue. Most people it seems encode at 128kbps using some "fast"
encoder like Xing, which turns out something only marginally better
than modulated noise.
Now I'm no expert, but I imagine the main problem in designing a lossy
audio compression is gracefully handling distortion in the face of
reduced information. At the end of the day, if you are trying to get
the perceived quality of an ogg-vorbis file at a quarter to half the
size, then you have a lot less information to play with in your
reconstruction.
As an aside, I've always wanted to try and factor out the inherent
redundancy in music. ie: if there's a repeating drumloop, subtract it
from the audio and encode it seperately. Of course, this would need to
operate on large blocks of wave data (ideally the entire input), and
be incredibly slow.
> If a new algorithm for audio compression would emerge, would it not take a
> long time before people replaced their mp3's with the new format?
> Considering the size of current hard-drives and the steadily increasing
> internet bandwidth, how good (in terms of compression-rate, quality
> (subjective/objective), decoding/encoding speed) would a new audio
> compression algorithm have to be to be considered useful for the masses. Are
> there any niche-markets in which such an algorithm would be useful?
Another problem is the large installed base of mp3 applications.
Switching formats is generally a pain... Most of my albums I own I've
got encoded as mp3, although all the stuff I've bought over the last 9
months or so I've used ogg vorbis to compress. I use the same average
bitrate as I did with mp3, just the quality is higher. On the whole
I'm happy with the shift, although my car can only play mp3, so I have
to convert my newer stuff if I want to listen to it in there. If you
bring out a whole new format, it wont really take off until everyones
favorite encoder frontend & player software supports it.
- Geoff T
Thomas Richter
> If I were to invent an audio compression algorithm that could compress audio
> to half the size (or a quarter the size) of e.g. mp3 or ogg-vorbis with the
> same psycho-acoustic quality, how useful would it be?
I'd say that this sounds quite useful. Consider satelite audio transmissions,
e.g. long distance phone calls and similar.
> Assume that the
> computational cost of decoding/encoding is 2-4 times that of mp3 or
> ogg-vorbis. Are there any uses for high-compression and "slowly"
> encoded/decoded audio compression algorithms?
Yes.
> If a new algorithm for audio compression would emerge, would it not take a
> long time before people replaced their mp3's with the new format?
This very much depends on how much better the algorithm is. For a new
algorithm, it really requires a "quantum jump" to get some acceptance
by the general audience. Thus, if it is 10% better, then possibly no. If
it can compress twice as good or better, then possibly yes.
> Considering the size of current hard-drives and the steadily increasing
> internet bandwidth, how good (in terms of compression-rate, quality
> (subjective/objective), decoding/encoding speed) would a new audio
> compression algorithm have to be to be considered useful for the masses.
Hard to say, and this is not the only factor. Make your compression standard
open, standardize it, offer free decompressors, and you'll have much better
chance.
> Are there any niche-markets in which such an algorithm would be useful?
Sure, any kind of application offering only a low bandwidth.
So long,
Thomas
Gilles Vollant
I believe the WMA codec from latest Microsoft media player offer better
compression than MP3
Geoff T
> I think this will be useful.
>
> I believe the WMA codec from latest Microsoft media player offer better
> compression than MP3
Quite possibly... MP3 has been around for quite a long time. I
remember when the first WMA codec came out - some of my friends
converted their entire 128kbps mp3 collection to 64kbps wma. I
resisted, mainly because I listened to bands that *didnt* record
inside 44-gallon steel drums.
- Geoff T
Dcoder
It offers the same quality of mp3 at 128 kbs in just 48kbps! It's awsome!
The encoder comes with nero 6, and i can get a winamp plugin to play this
format.
"John Reidar Mathiassen" <joh...@itk.ntnu.no> wrote in message
news:biavos$9vs$1...@tyfon.itea.ntnu.no...
Thomas Richter
> Have you heard of HE AAC?
> It offers the same quality of mp3 at 128 kbs in just 48kbps! It's awsome!
> The encoder comes with nero 6, and i can get a winamp plugin to play this
> format.
While I believe that you can get better than MP3, it seems doubtful to me
that you can get that much better without paying an (audible) quality price.
So long,
Thomas
menno
48 kbps may be a bit optimistic for the trained listener, but at 64kbps it
is comparable to 128kbps MP3.
The trick lies in the fact (first of all AAC without the HE extension is
alread a lot more efficient than MP3) that the high frequency part of the
audio is not "encoded" but parametrised and stored using just a few
parameters. Try it :)
Menno
tal...@nospam.wp.pl
audio
> to half the size (or a quarter the size) of e.g. mp3 or ogg-vorbis with
the
And is faster than mp3. I think your product must compress better than wma
to be sold.
Regards,
Talthen