On 2013-11-05, Eero Aro wrote:
The previous posters already said it all.
Actually, not all of it. Evenif I drafted a way to find it all, I *anything* but sure it would work out in the end.
I can only add that it is almost impossible to find out from an acoustic Soundfield or other microphone setup recording, whether it is UHJ or not.
Eero, using modern digital processing, I'm pretty sure it's not. On the contrary, I'm pretty sure it could be regularly be found out. And in excess of that, over the longer term, it ought to be possible to find out which precise variant of BHJ they used to begin with. Because there are at least three.
It is sometimes possible to see from a goniometer if a multitrack recording from mono microphones has been encoded into UHJ.
Yes. But then you'd be working the short term. If you did you discering in the longer terms, and in a statistically principled fashion, would you even then say say the same thing? I don't actually think you would, speaking as a mathematically/statistical-recognition/supervised-learning kinda guy.
Or actually - you can see the difference between a "normal" panned stereo and UHJ, as the directions of individual panned sources can be detected from "normal" stereo.
They can, in each encoding, separately. But how precisely do they show up? Especially after most of them encodings prove old analog ones? Coming from old analogue tapes? Seriously, even if you selfevidently *can* detect and decde the stuff, blindly, what you need is some time and some rather heady DSP. Something quite beyond the normal LTI-stuff, even.
They look like sharp(ish) direct lines, but UHJ doesn't have such components anywhere else than on the edges of the two speaker stereo, ie. at -45 and +45 degrees. In all directions there are phase differences between the two channels.
The simplest, most robust thingy I know of on the detection side would be a full multiband Hilbert transform, which then goes bandwise (timewise and always lapped; you can actually do that in wavelet transforms too) to get both the imaginary and the real part of the signal out at the same time, even in a frequency separated fashion. Phase and all, and especially then easily calculating out the relative phase angles between multiple channels, both in and out.
By the way, did I not already tell you this? I might need some help in deriving an arbitrary second or third order decoder, which is good enough fr "any use". This time it's not even just theretical, but the real deal. Or as they say, "I suddenly cease to be a fuckker, and went primetime."
There that kind of detection might just be asked of me, robustly, for L/R vs. M/S-Dolby processed kind of thing. So, tell me more.
And if I now were to deliver, decoding wise, I'd have to do with something pretty novel, since ya'll just never, even before answered my questions about this sort of a decoder. "The fully general, yet not perhaps fully optimized one." ;)
Right now I'm rather terrified, at having to direct and/or actually do the coding work for something like this. Something like an arbitrary, at least second order decoder. So if you can help, please help as best you can.
-- Sampo Syreeni, aka decoy - de...@iki.fi, http://decoy.iki.fi/front +358-40-3255353, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
