Martin Geddes recently reflected on the use of Skype as a tool for recording podcasts with two people in different locations. This is a technique that is used on many podcasts now, including Blue Box, the VoIP Security Podcast. But as Geddes says, sometimes the quality is not all it should be, and it would be useful to be able to record in top quality, and in some way transmit this out-of-band, while using the inferior, real-time audio between the two podcasters. Sometimes this technique (called double-ending, or a “double ender”) is done manually today in podcasting and in radio: each person records their end of the conversation locally, then the files get spliced together at the end to make a broadcast quality programme. The telephone call only needs to be good enough for the two people to understand each other while the interview is taking place.
But adding double-ending functionality in Skype has interesting possibilities, apart from the podcasting one. In some areas human speech needs to be understood by less tolerant parties than humans, for example in the areas of automatic speech recognition, or speaker verification. Given that VoIP streams can be of cellphone quality (or lower), it could be useful for a computer system to be able to play back a passage of speech it was having trouble with. For example, a speaker verification system might listen to the live VoIP speech, perhaps match with a certainty of 20%, then after a few tens or hundreds of milliseconds it could try again using extra hi-fidelity information that came in while it was processing the first time. Much better than forcing the user to re-speak their passphrase over and over until the computer figures it out.
On the subject of Dan York (of Blue Box) and Martin Geddes, you can almost see them in this photograph from Fall VON. York is moving at speed, presumably in order to eclipse Geddes.