At 06:59 PM 10/20/02 -0400, Steve wrote:
>Am I missing something? If I can get decent bearings with one pair, and
>decent bearings with another (displaced) pair, won't the bearings intersect
>at the source of the sound?
>
>I was hoping to use my full complement (4) of microphones and record two on
>a DAT and two on a PC, synchronizing the sounds with clicker of the sort
>animal trainers use (I already got one for 59 cents, so don't tell me its
>hopeless!) Clicking at each mic would give all the inter-mic distances,
>enough to determine their relative locations, (assuming constant speed of
>sound).
>
>My plan was to try and chart the motions of a Red-eyed Vireo, say, over a
>fairly long time.
>
>Any hints, suggestions, instructions, corrections, comments from all you
>knowledgeable types would be greatly appreciated...
>
>Thanks,
>
>Steve P
This assumes both a life in a plane, and an ideal preservation of phase
relations as they arrive at each mic. The latter are notoriously screwed up
by digitizing the signals. MANY multi-channel digitizers actually strobe
from one input to the next, and perform a SERIAL or SEQUENTIAL
sampling. Later analysis usually presumes samples made at the SAME
TIME. This presumes a much higher degree of synchronicity among channels
than a clicker is going to give, I would bet.
It also implies that you know the physical layout of the mics with some
precision, as Walt has suggested. One would have to be able to input that
into the editor.
But I am far from an expert, let alone "a knowledgeable type". I am more
of a "hopeful", and perhaps a "somewhat experienced" type.
All this was once explained to me by Mike Collins, who, for "the Navy"
developed a retro-fitted computation scheme, called beam-forming by
simulated annealing, which works with arrays of, I think, four or more
mics. I think two mics ideally gets you a cone of bearing, and two more
gets a second cone, which if mics are exactly in a line, intersect on a
single line that wanders through space, but at slightly varying distances,
to get precise. You need another degree of freedom, I may be wrong, here,
another mic, to get it to a point, to produce A DISTANCE or pair of
distances, and six to resolve the remaining ambiguity.
But Mike Collin's algorithms are public domain, and my math and time
availability at this point are very limited. I would love you to add to
my previous posted list of ideal desired tools your editor could provide a
retro-computation system for location!
Mike's system actually back-calculated to produce a pure channel from each
separate source in the environment, based on bearing and distance. This
would provide a really superior noise removal, if and when we can implement
it! All the energy of noise would be allocated to its own point source
channel, and removed from each bird-specific channel.
I have referred to this study two or three times on this group, previously
over the years.
cheers,
Marty Michener
MIST Software Associates
PO Box 269, Hollis, NH 03049
coming next week : EnjoyBirds - software that migrates with you.
________________________________________________________________________
________________________________________________________________________
|