[edlea]

I can't work out if that would be possible or not.

It might be possible to plot the location of a single point in 2D, but how would you differentiate for multitouch? If you have 2 microphones you might be able to calculate the distance from each microphone of a touch by timing how long it takes for the sound of the tap to reach each microphone. However, as the speed of sound in wood appears to be ~4000 m/s (http://www.engineeringtoolbox.com/sound-speed-solids-d_713.h...) you'd need to be "listening" at 400kHz to get centimetre accuracy over a 1m square area.

I can't see how it would be possible to then differentiate between two points moving at the same time (i.e. a pinch gesture). I'm just assuming regular microphones are used as the article suggests - I'm not an acoustic engineer so hopefully I've missed something.

[ww520]

Interesting problem. Haven't thought of the sampling rate required. The location resolution probably will be poor for precise multitouch function. But even a single 2D location is enormously useful - turning any surface into a touch screen.

kabdib mentioned phrased array, which can be very helpful. It might solve the sampling rate issue since it's analog. I found this http://en.wikipedia.org/wiki/Beamforming. May be some of the techniques in radar/sonar can be borrowed here.

[kabdib]

Use a phased array of mics. Classic technology (well, 1960s).

Probably much harder with random mic placement.

[nitrogen]

Probably much harder with random mic placement.

You could run an initial calibration on the system by having the user tap a large number of points within the touch area and noting the relative phase, amplitude, and frequency content received by each mic for each location.