Help - Search - Members - Calendar
Full Version: Low Bitrate Codecs for a listening test
Hydrogenaudio Forums > Lossy Audio Compression > Other Lossy Codecs
MiVox
Hi,

for my diploma thesis I developed a codec for string instrument compression and in order to test it's performance I need to compare it to other codecs in a listening test.
What I need are about 5 different codecs I can compare it to. My Codec's bitrate varies from 1 to 7 kbit/s depending on the signal complexity. The competitors should range from this bitrate to maybe 30 kbit/s and they should use different compression techniques.

I'd be happy if anyone could give me some ideas.

Thanks!

MiVox
joeshrubbery
Those are very low bitrates indeed, assuming you're using standard CD quality audio. Off the top of my head I can only think of HE-AAC and maybe Ogg Vorbis as being even remotely viable, but they'll probably sound pretty bad even at the upper end of the bitrates you mention.
MiVox
Thanks for the answer. Sorry, I forgot to mention that I'm only working with mono signals by now.
And by the way, I didn't say my codec sounds good smile.gif

Are there no codecs intended for very low bitrate coding like those for speech?

MiVox
[JAZ]
I would suggest trying speech codecs in this case (or at least, not forget about them). Try speex (http://www.speex.org/), and you might want to experiment with GSM also.

You might want to compare to mp3 just "for comparison". ogg, and he-aac might be actual contenders, although they will be more on the 20ish range even for simple signals.

At last, you might want to test wavpack lossless codec aswell, but the size will be very dependant on the signal (being a lossless codec)
MiVox
Thanks for that. I'll take a look at these.

Are there any other non-speech very low bitrate codecs?

MiVox
Sebastian Mares
There was a special Vorbis encoder made by Garf IIRC that was able to encode to such low bitrates. Don't know if it's still available though. Maybe you can ask him?
Woodinville
QUOTE(joeshrubbery @ Jan 7 2008, 08:32) *

Those are very low bitrates indeed, assuming you're using standard CD quality audio. Off the top of my head I can only think of HE-AAC and maybe Ogg Vorbis as being even remotely viable, but they'll probably sound pretty bad even at the upper end of the bitrates you mention.


You can add WMA Pro 10 to the list of codecs that will be remotely viable. The latest test I know of HEAAC and WMA Pro 10 are neck and neck.

What I'm curious about is how your codec sounds vs. them.
MiVox
Thank you.

I didn't find a way to encode below 48 kbps with the WMA 10 Pro coder I found (dbpoweramp). Maybe there is a way, I'll keep searching.

HE-AAC v2 lead to ~16 kbps as I tried it on one test signal and that sounded quite good.

Actually the codec I worked on is not that interesting, since it only works on very special signals by now, i.e. one note after another played on a string instrument.

The main reason for the listening test is to find out if the artefacts of my codec (a sort of livelessness) are less annoying than those of other low bitrate codecs.
Speex for example produces lots of noise.

MiVox
muaddib
HE-AAC v2 uses PS which is needed only for stereo signal. Since you are using mono signals you should use HE-AAC v1.
MiVox
QUOTE(muaddib @ Jan 8 2008, 01:14) *

HE-AAC v2 uses PS which is needed only for stereo signal. Since you are using mono signals you should use HE-AAC v1.



I used the Nero encoder with a target bitrate of 10 kbps and didn't chose the codec manually. So I guess it encoded with HE-AAC v1 then.
I just heard about the improved performance of v2 at very low bitrates so I expected the encoder has chosen this version.



I also tried Ogg Vorbis now at 12 kbps ABR, I think this one will definately be in the test. The bitrate is comparable and it sounds totally different to mine.
lvqcl
QUOTE(MiVox @ Jan 8 2008, 12:29) *

I also tried Ogg Vorbis now at 12 kbps ABR, I think this one will definately be in the test. The bitrate is comparable and it sounds totally different to mine.


What is the sample rate of input file?
MiVox
QUOTE(lvqcl @ Jan 8 2008, 05:02) *

What is the sample rate of input file?



The sample rate is 44.1 kHz.

lvqcl
QUOTE(MiVox @ Jan 8 2008, 18:02) *
QUOTE(lvqcl @ Jan 8 2008, 05:02) *
What is the sample rate of input file?

The sample rate is 44.1 kHz.


For such low bitrates (12kbps) it is useful to resample input file to 16000 or 11025 Hz.
SebastianG
You may wanna try the MPEG4 HLIN codec which is one of the parametric audio codecs for extremely low bitrates. "HLIN" is short for "Harmonic and Individual Lines plus Noise".

You can download the final draft documents of the MPEG4 audio specification and the reference implementations from here. Though, I don't know whether the old reference implementation is any good. It's quite dated (seems to be 10 years old).

Cheers!
SG
MiVox
QUOTE(lvqcl @ Jan 8 2008, 08:42) *

QUOTE(MiVox @ Jan 8 2008, 18:02) *
QUOTE(lvqcl @ Jan 8 2008, 05:02) *
What is the sample rate of input file?

The sample rate is 44.1 kHz.


For such low bitrates (12kbps) it is useful to resample input file to 16000 or 11025 Hz.

Most of the codecs resample the signal at low bitrates so I think I would get about the same result if I would downsample in front. The 12 kbps .ogg file I got was downsampled to 8 kHz for example.

My codec works without resampling so it keeps those high frequencies but it has other drawbacks. So I'd like to keep at least this advantage besides the bitrate :-)
MiVox
QUOTE(SebastianG @ Jan 8 2008, 09:39) *

You may wanna try the MPEG4 HLIN codec which is one of the parametric audio codecs for extremely low bitrates. "HLIN" is short for "Harmonic and Individual Lines plus Noise".

You can download the final draft documents of the MPEG4 audio specification and the reference implementations from here. Though, I don't know whether the old reference implementation is any good. It's quite dated (seems to be 10 years old).

Cheers!
SG


Thank you. I'll try this one, the target bitrate seems to be in the range of my coder.

MiVox
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.