Friends:
You may remember that a while ago I was asking about using time of arrivale
differences to locate sounds. I've made a little progress since then and
thought I'd bring you up to date.
I tested out the routine (for my sound manipulation program) that computes
the bearing to a sound source based on difference in the time of arrival of
the sound at two microphones. I mounted microphones 63 cm apart on a meter
stick and began recording. I walked 5 steps away from the stick and clicked
a clicker. The I walked 2,4,6,10, and 18 steps parallel to the meter stick,
clicking at each distance. The results are in the table:
Steps Angle (deg.) Estimate (deg.)
0 0 0
2 22 19
4 29 35
6 57 56
10 63 60
18 74 72
So: I'm rather happy with the performance of the procedure. The recordings I
used were made outside, so there's wind, drips from snow melting (yes!) in
the trees, passing cars and airplanes in the recording as well as clicks. I
assume that pre-filtering to preserve frequencies characteristic of the
sound source would help with the problem of alignment in the presence of
noise.
The current method aligns the two tracks of a stereo recording by
considering the possible shifts of one relative to the other and computing a
"dot product" of sound sample values (divided by the norms of the sample
vectors). This is cheap/fast approximation of the correlation coefficient
that amounts to assuming that the average sample values are 0. The shift
producing the highest correlation between tracks is used to get the
difference dt in the time of arrival, and then the angle can be found as
arcsin( V dt/D) where V is the speed of sound in your planet's atmosphere
and D the distance between the microphones. I'll clean up the code,
implement the "real" correlation coefficient, and include this module in the
next release of Dora sometime in the next week or so.
My thought is to make three recordings using a similar set up in a field
where Song Sparrows and Indigo Buntings nest. I should be able to get the
location of a song and identify the individual (because they have distinct
songs). But: I certainly can't put out 3 DAT's and 3 pairs of Sennheiser
microphones (even the "cheap" kind I use). Do any of you have suggestions
about affordable equipment options that'd be appropriate in this setting?
Any suggestions, corrections, comments, etc. whould be appreciated.
Thanks,
Steve P
________________________________________________________________________
________________________________________________________________________
|