hi tapio,
On 01/21/2014 07:23 AM, Lokki Tapio wrote:
The SDM is a parametric method that differs significantly from
spherical harmonics based methods. It requires at least 4 omni mics
(we use 6) and the basic idea is as follows. An impulse response is
analyzed with a short time window and cross-correlations between mic
pairs are computed to obtain a time-difference-of-arrival (TDOA) in
each pair. With 6 mice you get 15 TDOA estimates. From these time
differences it can be estimated from which direction the sound energy
is reaching the microphone array. Such analysis is done for every
single sample (i.e. a short time window, 1ms, is shifted one sample
forwarded after the analysis) and the final result is a normal IR +
direction info (azi and ele) for every single sample. In the
reproduction, we take an IR from one omni mic (to get flat frequency
response) and distribute the samples of it to corresponding
reproduction loudspeakers. In other words, we convert one omni IR to
convolution reverbs for all reproduction loudspeakers. An d then we
convolve the anechoic music with these loudspeaker signals
(=convolution reverbs). In our current listening room we have 24
loudspeakers in a 3D setup. The accurate direction for all samples
can be obtained using VBAP, but it attenuates high frequencies a bit.
Therefore, we just associate every IR sample to closest reproduction
loudspeaker, which might introduce a small error in direction, but
the frequency response (and sound color) is untouched. With the SDM
method, the most authentic reproduction (of the sound in the concert
hall) in the lab can be obtained.
thanks for elaborating! very interesting method and survey!
one thing i'm asking myself: since you can only ever have one direction
per time window, does that mean that each direction estimate will
"trigger" a complete IR decay trail from the omni mic in that direction?
i.e. your directional resolution is really high, at the cost of
identical timbre of all reflections?
are there any perceptual drawbacks? i'm sure i wouldn't notice, unless
the carpet i'm standing on is knee-deep, but... :)
particularly, will the "seat dip" effect be averaged and spread into all
directions?
best,
jörn
--
Jörn Nettingsmeier
Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487
Meister für Veranstaltungstechnik (Bühne/Studio)
Tonmeister VDT
http://stackingdwarves.net
_______________________________________________
Sursound mailing list
[email protected]
https://mail.music.vt.edu/mailman/listinfo/sursound
