So someone has written a paper on how to listen to someone typing on a keyboard and convert that to the text that was typed. I read about at freedom to tinker, via Podz. The concept being that if you have a recording of the sound of someone typing for a longish period of time (15 – 20 minutes) then you can use machine learning techniques to figure out the text that was typed – taking into account the fact that different words, and letter-sequences have different frequencies in English. It’s not scary or anything — just interesting.
I have a friend who could, with some accuracy figure out the phone number that was “dialled” by listening to the key-press tones – all this reminds me of him.