[CELT-dev] Differences between low complexity and full complexity version of CELT?

[Gregory Maxwell]

On Wed, Aug 26, 2009 at 3:49 PM, raja gobi<raja.gobi at yahoo.com> wrote:
> I could not see any differences between the 2 graphs. I am sure I am missing something.

PEAQ is not always perfectly predictive of human listening tests.
The difference between low and full complexity is primarily the long
term predictor. PEAQ gives lower marks to the LTP than human listeners
do.

> Can you please comment on
>
> 1. What kind of source material will show a noticeable difference between low complexity and full complexity versions?

Speech and other highly tonal signals.

> 2. What are the main advantages with full complexity over low complexity overall?

Improved speech quality at low-bitrates.

> 3. Does the full complexity stream perform better with packet loss?

No. It performs worse with loss.