[Charlie X-Ray] is having some modern fun with the phone system by pulling dialed numbers from the audio track of YouTube videos (translated). The first step was to find a video where a telephone is being dialed and the sounds of the keypresses are audible. You can’t tell those tones apart, but a computer can. That’s because each number pressed generates a combination of two out of seven closely related frequencies. [Charlie] isolated the audio using Audacity, then wrote a python script to generate a spectrogram like the one above. By matching up the two dark nodes you can establish which two frequencies were played and decode the phone number being dialed. So how does this work again… find audio of a phone being dialed, decode the number.. profit?
