Tuesday, September 14, 2004
Speech Recognition on a Chip
A press release from Carnegie Mellon University indicates that they are undertaking a project to create a hardware-based speech recognition system. The primary benefit appears to be the speeds in which speech recognition can be performed in this way - "100 to 1000 times more efficiently than on a conventional computer".
Primary uses for this indicated in the article are speech-based interfaces for portable devices, and the ability for emergency workers to interface with complex databases hands free in the field.
Another interesting use, in my opinion, is to provide indexable text in real time for the constant recording of personal audio. At the very least it could provide a written record of everything I said, and map it to a timecode in the original audio/video record. If it could do this for all other speakers nearby, all the better.
It's also a step closer to a universal translator, but more on that later.
Carnegie Mellon engineering researchers to create speech recognition in silicon