Home > Archive > Compression > November 2006 > Lossy Compression of Audio Noise
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Lossy Compression of Audio Noise
|
|
| Oliver Wong 2006-11-14, 6:55 pm |
| I'm having an argument with someone, and I can't seem to convince them
of my position, so I thought I'd s help on establishing a convincing
explanation for that person from this newsgroup.
We were talking about a voice-chat feature for a massively multiplayer
game, and eventually the other person said that noise is uncompressible.
I responded that in the context of voice chat for a game, we could use
lossy compression in which case it's trivial to compress noise, and told
them to google psycho-acoustic analysis.
They then told me to look at Shannon's definition of information
entropy, and I tried telling them that Shannon's definition only applies to
lossless compression schemes, and I gave them a proof of concept codec for
lossily compressing pure noise:
Encoder: Record the volume and duration of the noise.
Decoder: Generate random noise at the given volume for the given duration.
They then said that I'm not "really" compressing, I'm actually throwing
the information away and recreating it.
I told them in the context of lossy compression, that's what ALWAYS
happens. JPEG is "throwing away" the original image, and recreating a
similar, but different image later on. So does MP3 or Ogg Vorbis.
They say there's a difference between compressing and
throwing-away/recreating information, and I told them
throwing-away/recreating information falls under the generally used
definition of lossy compression, citing Wikipedia:
http://en.wikipedia.org/wiki/Lossy_compression
<quote>
A lossy data compression method is one where compressing data and then
decompressing it retrieves data that may well be different from the
original, but is "close enough" to be useful in some way.
</quote>
Now we're at an impass. So I was wondering if I could try to resolve the
argument by asking about it here.
- Oliver
| |
| Jim Leonard 2006-11-14, 6:55 pm |
| Oliver Wong wrote:
> We were talking about a voice-chat feature for a massively multiplayer
> game, and eventually the other person said that noise is uncompressible.
Pure random noise *is* impossible to compress -- LOSSLESSLY. But voice
chat programs use lossy compression.
I'm astonished the both of you are getting hung up over the definition
of lossless vs. lossy compression. If your friend doesn't want to
think of lossy algorithms as "compression", then he's free to do so,
however he'll get into this argument all the time since both methods
are generally considered compression techniques.
The Wikipedia article http://en.wikipedia.org/wiki/Data_compression
mentions both techniques as classifiable as "compression".
So what's the actual dispute?
| |
| Oliver Wong 2006-11-14, 6:55 pm |
| "Jim Leonard" <MobyGamer@gmail.com> wrote in message
news:1163547706.280170.308510@b28g2000cwb.googlegroups.com...
> Oliver Wong wrote:
>
> Pure random noise *is* impossible to compress -- LOSSLESSLY. But voice
> chat programs use lossy compression.
>
> I'm astonished the both of you are getting hung up over the definition
> of lossless vs. lossy compression. If your friend doesn't want to
> think of lossy algorithms as "compression", then he's free to do so,
> however he'll get into this argument all the time since both methods
> are generally considered compression techniques.
>
> The Wikipedia article http://en.wikipedia.org/wiki/Data_compression
> mentions both techniques as classifiable as "compression".
This is exactly what I suspected. Thanks for confirming I wasn't going
crazy. ;)
>
> So what's the actual dispute?
Literally, the dispute was as I've laid it out in the original post. At
a higher abstraction level, the problem is simply that neither of us will
admit that they're wrong in this particular argument. In my case, I truly
believe I am correct. As for him, I don't know if he actually believes in
what he's arguing, or if he realizes he's wrong, but just unwilling to admit
it.
- Oliver
| |
| Thomas Richter 2006-11-15, 7:55 am |
| Oliver Wong wrote:
> I'm having an argument with someone, and I can't seem to convince them
> of my position, so I thought I'd s help on establishing a convincing
> explanation for that person from this newsgroup.
>
> We were talking about a voice-chat feature for a massively multiplayer
> game, and eventually the other person said that noise is uncompressible.
>
> I responded that in the context of voice chat for a game, we could use
> lossy compression in which case it's trivial to compress noise, and told
> them to google psycho-acoustic analysis.
>
> They then told me to look at Shannon's definition of information
> entropy, and I tried telling them that Shannon's definition only applies to
> lossless compression schemes, and I gave them a proof of concept codec for
> lossily compressing pure noise:
>
> Encoder: Record the volume and duration of the noise.
> Decoder: Generate random noise at the given volume for the given duration.
>
> They then said that I'm not "really" compressing, I'm actually throwing
> the information away and recreating it.
>
> I told them in the context of lossy compression, that's what ALWAYS
> happens. JPEG is "throwing away" the original image, and recreating a
> similar, but different image later on. So does MP3 or Ogg Vorbis.
>
> They say there's a difference between compressing and
> throwing-away/recreating information, and I told them
> throwing-away/recreating information falls under the generally used
> definition of lossy compression, citing Wikipedia:
> http://en.wikipedia.org/wiki/Lossy_compression
> <quote>
> A lossy data compression method is one where compressing data and then
> decompressing it retrieves data that may well be different from the
> original, but is "close enough" to be useful in some way.
> </quote>
>
> Now we're at an impass. So I was wondering if I could try to resolve the
> argument by asking about it here.
Possibly the wording is just imprecise. There is "redundancy reduction"
where you reduce the bandwidth by removing redundant information of a
source, and there is "irrelevance reduction" which removes irrelevant
information from a source. The former is the world of Shannon
information and lossless compression, the latter the world of lossy
compression and modeling.
So long,
Thomas
|
|
|
|
|