Massachusetts Institute of Technology: Speech Recognition
At the MIT Media Laboratory, the Speech Research Group is conducting research in a voice-interactive user interface for a hand-held computer.
According to project leader Chris Schmandt, the group is developing new input and output methods that use speech recording, playback, recognition, and synthesis, since the value of larger I/O devices (keyboards, mice, and displays) decreases as computers continue to shrink.
Current research focuses on two areas: exploring the usefulness of voice as a “data type” such as text or graphics; and the extent of the role of voice-recognition software in the user interface.
The prototype interface lets users create and manage lists of “voice notes,” which are digitized and stored segments of the user’s speech. Users can “tag” individual segments of speech by speaking a name. Speech recognition software then translates the name and files the segment, which can be accessed later (also using speech).
The Speech Research Group is also experimenting with compression software for voice data, to let users scan and browse stored data, listening to shorter segments of speech.
Contact Chris Schmandt, Speech Research Group, MIT Media Laboratory, Massachusetts Institute of Technology, Cambridge, MA 02139; (617) 253-1000.
