Robotics Reference
In-Depth Information
31.1m
23.1m
11.6m
8.6m
8.4m
7.9m
7.6m
6.3m
0.75
straight
0.60
Figure 34.
Scene interpretation by the EyeQ chip (Courtesy of Amnon Shashua
12
)
Lip Reading
As more of the technologies within the field of AI become mature, so
there will be an increase in intelligent systems that rely on more than
one AI technology. An example of the power of hybrid AI systems is the
use of lip reading technology, based on computer vision, to increase the
accuracy of speech recognition systems. The technology of automatic
speech recognition has been developing steadily, albeit slowly, for more
than 50 years, to the point where the best systems work reasonably well
for tasks such as office dictation. But speech recognition is not effective in
noisy environments, and the technology may stumble for other reasons.
13
Teaching computers to read lips is one way to increase the accuracy of
automatic speech recognition, to help them understand the difference
between “bat” and “pat”, for instance.
Chalapathy Neti, a senior researcher at IBM's Thomas J. Watson Re-
search Center in Yorktown Heights, is using face recognition and similar
technology to boost the performance of speech recognition, in the same
12
Professor Shashua is chairman of the School of Engineering and Computer Science at the He-
brew University of Jerusalem.
13
See the section “Speech Recognition” later in this chapter for a mention of some of the factors
that can affect the accuracy of that technology.