Parametric Stereo, I am interested.......... Options

[iwod]

Consdiering HE AAC was very good already. I wonder what benifit does it bring.

And to what i read it seems to only provide benifits to low bit encoding.... As said on the site 24Kbps........

SO is this avalible on 48 or 64? Or is it limited to 24Kbps?

[menno]

The PS tool for AAC will only work together with SBR (HE AAC). So it will only be useful for low bitrates. Note that this is another parametric tool, like SBR, meaning that it can reconstruct an estimation of the original stereo image, it's not meant for complete transparancy at higher bitrates.

Another advantage might be the complexity. I'm not exactly sure yet how much processing this tool takes, but definately not very much (uses SBR filterbanks, which are done already anyway). However, the core codec (AAC+SBR in this case) will only have to decode 1 channel instead of 2 (50% gain). Only the SBR synthesis filterbank still has to be performed twice.

Menno

[Garf]

If I read it correctly, it means that PS actually could lower processing cost of HE-AAC playback?

[menno]

If I read it correctly, it means that PS actually could lower processing cost of HE-AAC playback?

Yes, unless it takes more processing power then 1 channel of AAC+SBR, but that is very unlikely

Menno

[AgentMil]

Nice... look like AAC is the codec to beat in the future. Defintely look at purchasing an AAC enabled portable player now.

Regards

AgentMil

[iwod]

ok, so in layman terms, it is a Enhanced SBR......

But.... by the sounds of it, it looks like this PS is for seriously low bitrate. i.e the example 24kbps.

Because i don't see how this could fit in to 48 or 64 kbps..........

The lower of processing power in theory sounds good. Although i doubt there wll be much difference in pratice.........

Sorry for my bad is SBR synthesis filterbank performed in Real time?

And if i am correct, PS is just a better way to try and recontruscting the orginal signal.........

This post has been edited by iwod: Jan 31 2004, 10:41

[menno]

ok, so in layman terms, it is a Enhanced SBR......

Not really. It's basically a whole new codec on top of AAC+SBR. It just makes use of the SBR filterbanks.

But.... by the sounds of it, it looks like this PS is for seriously low bitrate. i.e the example 24kbps.

Because i don't see how this could fit in to 48 or 64 kbps..........

I suspect it will be useful at bitrates below 48 kbps. There will probably be a turning point where AAC+SBR is better then AAC+SBR+PS, just like there is a turning point where AAC is better then AAC+SBR.

The lower of processing power in theory sounds good. Although i doubt there wll be much difference in pratice.........

Sorry for my bad is SBR synthesis filterbank performed in Real time?

Can you playback HE AAC in realtime? Yes

If this tool would mean 40% lower processing power, it means that AAC+SBR+PS will use about the same amount of processing as stereo LC AAC.

Menno

[David Nordin]

I'm very much looking forward to this being properly implemented for testing. Good news indeed.

[21_already]

I know it's part of one of the Mp4 profile, but who made it?

Is it going to be applied to any other codecs like SBR is applied for MP3pro ?

With a codec like this from the sound of the description it seems like it could pull individual sounds out from a sample, so like in a song it could pick apart vocals from guitars, the idea being that coding sounds (or types of sounds) individually takes less data then coding them bundled all together... or is this complteley wrong?

One last question:
Will Nero implement it anytime soon ???? Pleeeeeze

[Dibrom]

or is this complteley wrong?

Yes.

[menno]

I know it's part of one of the Mp4 profile, but who made it?

I think it's a joint development between Philips and CodingTechnologies (I think mainly Philips, since CT uses a different, but somewhat comparable, PS system (in DRM for example)).

Is it going to be applied to any other codecs like SBR is applied for MP3pro ?

MP3Pro already has a tool like this called LC_Stereo. I think it gets used at extremely low bitrates only (you can test that by decoding it with a normal mp3 decoder). Anyway, this is way less advanced then the PS tool developed for HE AAC, it uses some simple filtering.
However something like this could be used on different codecs. It will be used in the mpeg-4 parametric codec, for example.

With a codec like this from the sound of the description it seems like it could pull individual sounds out from a sample, so like in a song it could pick apart vocals from guitars, the idea being that coding sounds (or types of sounds) individually takes less data then coding them bundled all together... or is this complteley wrong?

Unfortunately you are completely wrong, the first person that solves the so-called cocktail party problem will probably receive a Nobel prize right away

One last question:
Will Nero implement it anytime soon ???? Pleeeeeze

Drumrolls...... you'll see

Menno

This post has been edited by menno: Feb 2 2004, 19:58

[ProtectYaNeck36]

is parametric stereo somewhat of a pseudo-stereo immplementation for aac at low bitrates?

[Ivan Dimkovic]

Something like that - basically, you only code mono signal with HE-AAC codec, and spatial-cues separately.

Think about it as much improved Intensity Stereo

- Intensity stereo only records Interaural Intensity Difference (IID) while Parametric Stereo stores much more relevant data (Interaural Phase Difference, Interaural Cross Correlation, etc...) - while IS destroys phase information, PS stores it as well.

- Intensity stereo operates inside codec T/F filterbank (256/2048 samples in AAC) which makes it non-optimal in case of fast stereo image changes like with transients. PS uses separate T/F grid making much better T/F maping possible.

Also, PS encoding takes less time than encoding of stereo HE-AAC.

[LordofStars]

With this addition I will completely give up use of other codecs. I only use mp3 to use on my portable, but with something IS like I could definately validate buying a aac portable. HE-AAC itself allowed me to justify purchasing a nero liscence.

[julzhu]

Could anybody provide me a link of a tool to generate enhanced AAC plus file,
and how this file header will be?

Thanks
Julzhu

This post has been edited by julzhu: Oct 6 2004, 23:01

[wkwai]

Something like that - basically, you only code mono signal with HE-AAC codec, and spatial-cues separately.

Think about it as much improved Intensity Stereo

- Intensity stereo only records Interaural Intensity Difference (IID) while Parametric Stereo stores much more relevant data (Interaural Phase Difference, Interaural Cross Correlation, etc...) - while IS destroys phase information,  PS stores it as well.

- Intensity stereo operates inside codec T/F filterbank (256/2048 samples in AAC) which makes it non-optimal in case of fast stereo image changes like with transients. PS uses separate T/F grid making much better T/F maping possible.

Also, PS encoding takes less time than encoding of stereo HE-AAC.

A question about PS.. What is the range of frequencies PS is applied on?
For the IS, it isn't recommended to apply IS on frequency bands below 8 Khz
because of the sensitivity of the human ear to phase information.. I wondered if PS would address this issue ?

wkwai