Voice I/O still remains an elusive "holy grail" of user
interfaces. The voice output part is getting very good. Voice synthesis
has reached a point where most people can understand it and techniques
for splicing human voice segments together sound quite natural. There
have also been many accumulated lessons learned on when and were to
use voice I/O so it is not annoying. Voice input still remains a difficult
problem. However, a few years ago, Lernout & Hauspie, a Belgian
company, developed a set of software products that not only did voice
input and output better than anyone had before. |
|
But solved
most of the problems of speaker independence (recognition without
training), and connected speech (the ability to parse words out
of a continuous sentence). L&H voice I/O technology is used
by Home Automation Lab's HAL line of software-based HA system products.
The company quotes an 80-90% recognition rate from any untrained
speaker on the first try as long as the user is willing to speak
fairly standard commands.
HAL software-based home automation system uses voice input and output
as the main user interface. The system can recognize requests from
most speakers without any training.
The other more practical problem of voice I/O is getting the voice
output where it is needed and having microphones available to capture
voice input. HAL's product, as well as other voice input products,
use the telephone as the primary voice input and output device although
systems can use a set of microphones and speakers placed at strategic
locations in the home. |
| |
|
