Jag har sökt men inte hittat det inlägg där ingvar diskuterade lite om sin teknik at spela in stereo med höjdinformation, oh jag hitttat det skull jag skrivit detta inlägg i den tråden.
Nu är situationen som sådan att jag har gjort ett inlägg i
sursound / ambisonic mailinglistan som fick till resultat att
forskaren Sampo Syreeni skulle vilja utnyttja ingvar's forskning.
I detta fall de olika generella frekvens modifieringar som genererar
den psykoakustiska effekten av höjd.
Jag fick uppfattning att dessa är resultatet av en stor undersökning av hörseln som Ingvar gjort.
Eller har det redan presenterat i någon artikel som kan refereras till?
Detta är ju inte ingvars privata mail men du / han läser och skriver mera aktivt här än i sin mail ':P'
Hoppas det kan förlåtas, lite allmän kunskap kan kanske fås i alla fall.
En del av termerna som finns kan förstås om man läser mera på websiten
www.ambisonia.com.
Här kommer de mail som orsakar detta inlägg:
On 2008-03-14, Bo-Erik Sandholm wrote:
> A "phsycoacoustician" (Ingvar Öhman) that I am aquainted with,
> designed and implemented a way of recording stereo with height using 2
> microphones with a varying frequency response depending of heigth.
> Spaced omnis was the basic microphone placement.
This is highly interesting to me. Obviously such a system is inferior to full, periphonic ambisonic. But I wonder whether the impulse response functions resulting from the research you mentioned could be used a) to yield some isotropic simulacrum of height using a horizontal ambisonic rig only and/or b) to suggest a fully invertible means of coding height into G-format which b1) also sounds reasonably good/realistic without being decoded.
That'd be the ultimate UHJ analog for G-format, as somebody suggested a few months ago when I (once again) talked about compatibility formats.
--
Sampo Syreeni, aka decoy - mailto:decoy@iki.fi, tel:+358-50-5756111 student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2
--------------
From Stefan Shreiber:
I have thought about this (height coded in "G-format") before, although on a very general level.
I think a "G format" with height won't be feasible, because every height information won't be recoverable. At least not completely.
Underspecified coding/decoding process?! (You might be able to have height if you are doing a sub-optimal B format to 5.1 encode?)
In classical G format, you won't use < any > height information in the
5.1 track. At least this is my understanding.
But all this is certainly for a serious mathematician, so I give the questions on...
Interesting project, but shouldn't work, for "intuitive" reasons.
At least, I don't see any "evident" way.
Best,
Stefan
P.S.: I guess we are not the first ones here...
If it could be done, somebody would have done it?
--------------------
On 2008-03-15, Stefan Schreiber wrote:
> I think a "G format" with height won't be feasible, because every
> height information won't be recoverable. At least not completely.
I'm not suggesting a binaurally optimized G-format, or anything like.
I'm saying, there are two degrees of freedom available within the G-format signal set after encoding a 1st order signal set in any way at all. So, why not first a) sum some signal derived from Z into the 5.0 channels in a recoverable form, and then b) try to optimize that signal towards a psychoacoustic decode that'd at least partially simulate height, while retaining perfect recoverability?
What I suggested earlier was to use phased sums and differences to embed Z, let's say by by phasing the signal to +/- 45/135 degrees and adding it to the main channels at 1/4th the amplitude. Some smoothed version of the HRTF analysis could then be used to modify the spectral balance of the input signal such that it resembles the cues that would be given by an up signal. That's clearly the lowest denominator and not psychoacoustically tested and/or optimal, but it could serve as a starting point.
> In classical G format, you won't use < any > height information in the
> 5.1 track. At least this is my understanding.
Quite correct. But you know, that is *also* the reason why those extra two degrees of freedom stay unused, to be used for other purposes. As long as they don't sound *too* bad when utilized in the context of the original horizontal decode.
Earlier I also suggested one might be able to encode the center front channel into the last remaining degree of freedom. Perhaps that's a good idea, perhaps not. But I think even that is worth the shot unless we can show that height requires both of the degrees of freedom in any and all intermediary/transfer formats.
--
Sampo Syreeni, aka decoy - mailto:decoy@iki.fi, tel:+358-50-5756111 student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2
