Use inside my opus library about rust_dct

[ejmahler]

I'd be happy to implement those two. The only reason I haven't so far is lack of information on how. Wikipedia and other easily available sources only have definitions for the simplest case, which is what I've implemented. If you have some resources on how to implement arbitrary overlap/stride I'd gladly look into it.

[est31]

I've done a little bit of research and it seems the stride is something connected to a rotation, which is specific to opus, and probably done in the same function for performance reasons... I guess its not the full rotation but only some preliminary stuff but I need to consult the function more closely.

Overlap seems to be that it only overlaps the ends of two windows, and not up to the half or something. I also think that opus may have windows with different sizes following each other.

[ejmahler]

After reading, I don't think the rotation is related to stride. I think the rotation happens earlier on, before the IMDCT is involved. In section 4.3.7 a few pages down, it says this:

The inverse MDCT implementation has no special characteristics. The
input is N frequency-domain samples and the output is 2*N time-domain
samples, while scaling by 1/2.

Aside from the scale factor of 1/2, the MDCT library in this crate does exactly what the paper describes.

[ejmahler]

Closing this due to inactivity. If you think there's still something to be done here, feel free to open a new issue

[est31]

Sure! I've more or less stopped working on the library anyway, and am waiting for someone else to have finished an opus library.