Forums

DCT in audio compression

Started by eblade September 20, 2008
After reading, what seems like a lot of material, I'm still in doubt why
DCT isnt used in audio-compression. Seeing as DCT assumes mirror symmetry,
my only idea for not validating DCT over MDCT is that with real-life
signals you can rarely (if ever) expect a signal to have that property? Is
that correct or is there a better explanation.


eblade wrote:

> After reading, what seems like a lot of material, I'm still in doubt why > DCT isnt used in audio-compression. Seeing as DCT assumes mirror symmetry, > my only idea for not validating DCT over MDCT is that with real-life > signals you can rarely (if ever) expect a signal to have that property? Is > that correct or is there a better explanation.
Never having tried it, I think it might be too audible. I am noticing more and more the little squares in digital television, especially on fast changing scenes. (When I first noticed it on the olympic logo going in and out of a commercial break I thought it was a new special effect. Later I decided it was an artifact.) The rules for what is inaudible and can be removed are somewhat more complicated, but at the lower data rate you can do a lot more processing. -- glen
>After reading, what seems like a lot of material, I'm still in doubt why >DCT isnt used in audio-compression. Seeing as DCT assumes mirror
symmetry,
>my only idea for not validating DCT over MDCT is that with real-life >signals you can rarely (if ever) expect a signal to have that property?
Is
>that correct or is there a better explanation. > > >
Having read a little more, MDCT reduces spectral leakage by multipling a Hann window (or some other sine-window). Does that mean that DCT doesnt have spectral leakage?