> [A quick contemplation of the wavelength of the sounds in question
> would put an end to that speculation I suspect. --Perry]
I know this has been somewhat done to death, but there's a nice
comparison: GPS positioning using carrier phase tracking is equivalent
(well, it's reversed - clicks come from the microphones/satellites and
the key/receiver calculates its position - but the principle is the
same). This can give millimetre accuracy with carrier wavelengths of
19cm (if you're very careful, have lots of time and maybe some luck).
The precision comes from cross-correlating wave trains rather
than trying to measure a particular point (eg the initial rise of the
click) accurately. You wouldn't do as well with keyboard clicks, but
then you don't need to.
Note that usually GPS positioning is not done using carrier phase
tracking - that, together with problems like different atmospheric paths
from differnet satellites and, in the past, noise added to civillian
signals, gives much lower precision. See, for example,
http://www.colorado.edu/geography/gcraft/notes/gps/gps.html
Accuracy for keyboards would depend on how many wavelengths can be
detected at good signal-to-noise within a single "click" (and having
stable recordings with no wow or flutter). Also, it would be useful to
know the identity of one key - return for example - to help solve for
the position of the keyboard relative to the microphones. Getting an
initial solution might be difficult - it would be a big help to know the
relative position of keyboard and mcirophones to within a wavelength or
two (and have all recordings marked by synchronized clock ticks). If
the user moved their keyboard during typing it would cause havoc with
any attempt to converge on a solution. Maybe we should all start
walking around as we type...
Andrew
(In a previous job I wrote software to calculate positions from GPS
satellites - Paul Crowley may be able to correct me if I have made any
errors as he was there too...)
