Sunday, October 20, 2013

Another Astonishing Automation

From, 2009/07
So... How will Star Trek: The Next Generation tech affect your job a a BA?

Summary Article: Automatic speaker tracking in audio recordings (Science Daily, retrieved 2013-10-23)

Years ago it was clear that human speech would be completely comprehendible to computers. Now, a team of researchers has made a big advance in the field. It is one thing to interpret one person in a quiet room. Several people in conversation is a different task entirely.

Part of my joy in reading this on a sleepy Sunday comes from the realization that there is an entirely different kind of geometry that I had never heard of.


Journal Reference:
  1. Stephen H. Shum, Najim Dehak, Reda Dehak, James R. Glass. Unsupervised Methods for Speaker Diarization: An Integrated and Iterative ApproachIEEE Transactions on Audio, Speech, and Language Processing, 2013; 21 (10): 2015 DOI: 10.1109/TASL.2013.2264673

No comments: