On 2021-03-05, Ralph Glasgal wrote:
Checkout the AES papers at www.ambiophonics.org. But basically if you have a front pair that was recorded with accurate values of ITD and ILD and a rear pair isolated from the front that also has accurate ITD and ILD then you can have a full circles of very realistic sound in the horizontal plane using RACE, BACCH or similar crosstalk cancellation apps with just four speakers.
I've been down and under for quite the number of months now, or even a couple of years, thanks to COVID-19. Sorry. As such, let me return into the fray for a while?
Binaural to FOA is an underdetermined problem, mathematically speaking. It's two nasty channels to four straight. That means that if you want to somehow "solve" the inverse problem, you will have to bring in hard a priori knowledge.
Statically you cannot have that knowledge. If the ears stay put, it remains a linear underdetermined problem.
However if the ears move, you can posit the problem as a superresolution one. In space/orientation. Any steady sound source can be taken as a probe of the binaural soundscape, and the HRTF. You *can* in fact do statistical inverse MIMO analysis of if, as long as you have some a priori knowledge of the signal's true statistics.
That means you *do* tend to be privy to some a priori information, towards the solution of the binaural-to-FOA (HOA?) problem. It's just that you'd need to know or infer which way the ear-pair were, from auditory analysis alone, given a binaural reference, found out by automation.
It *can* be done. It's only that the inverse problem is a highly dynamical one, and one which hasn't been solved, nay, even attempted at yet.
_______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound - unsubscribe here, edit account or options, view archives and so on.
