Interesting (from a technical point of view). I once had to look at how to encode timestamping in transmitted audio for sync purposes (long story). The method that was least perceptible was phase shifting frequencies (or select frequencies) at a period below 60Hz. Anything faster and it may be perceived as a sound. The other problem is that the transmission of most ‘voice’ speakers is above 200Hz and below 3kHz. While some people may have high quality audio systems attached to their TVs, not everyone does. I suspect that the timestamp pulse is more like every minute or so, or even less to avoid collision between transmissions. It could be that a encoding phase transitions occur every 2-3 seconds – kind of a 1/2 baud transmission.
You are kind of doing audio-steganography. If the bandwidth of the data being transferred breaks into the region of human hearing – it will be perceived.
Now I am tempted to digitize a bunch of network audio – and find out if I see phase shifting… I wonder if I can see the phase shifting in digitally encoded audio from video data – some video encoding uses something similar to mp3 encoding for audio.
