Home

when adam was talking (in one of his daily source codes ) about adding meta data to (mp3) sound files he was thinking about the mp3 id-tags but the following idea came to my mind:

in-band signaling

you could send some tones that represent data in the audio itself.

there’s a few different standards that already describe something like this so first i thought about the way the deaf people have text telephone but that is a modulated signal (just like a modem) and basically it sounds horrible.

but i reckon just the keypad tones that a phone uses will be fine.
Dual Tone Multi Frequency DTMF uses 2 tones to pick a number (or * or # or A, B, C or D)

so to encode short bits of information like URL’s in your audio you could start sending ascii codes delimited by for example the hash key

so

would be encoded as

try how that would sound in your mp3 here.
(just a note, this applet seems to grab hold of your key input, so just
close the window once you are done

i know it takes a while but i’m sure you can speed it up and it will still be recognisable..

then on the other side we need an application that sits and listens to all audio.
with a bit of digital signal processing (i think they use the Fast Fourier Analysis) the individual
characters can be extracted and shown in a window as the tones are ‘received’..

if this application finds a URL it can even start up a browser and navigate to it.

if the sounds is too annoying just add them to the end of your podcast so people that don’t know how to handle it can end it right there. this way you keep your podcast and the meta data together and you don’t need to invent another new standard.

the generation is the easy bit, the detection part is a bit harder to do.

just an idea.. any comments anyone ? is there a better protocol (one that actually already has all characters encoded for example) already out there ?