NSF PR 01-107 - December 17, 2001
New Public-Domain Database Could Advance Human
Computer Interaction Through Sound
Researchers in California have created a new, publicly
available database of acoustic measurements of human
subjects that may help engineers build personalized
sound systems for computers that could rival or even
exceed the experience of listening to a high-end home
theater system.
Richard Duda and V. Ralph Algazi of the University
of California, Davis said the database could have
a wide range of applications, including teleconferencing,
mobile computing and home entertainment. The National
Science Foundation (NSF) funded their work.
"One day," said Algazi, "computer users could operate
a small, 'wearable' computer using voice commands,
with spatial sound replacing a visual display." He
added that the database could aid in the development
of "immersion" systems that could allow scientists
to interact with their data in a computer-generated,
three-dimensional space incorporating both images
and sound.
People use a number of complex sound cues to experience
their surroundings. But reproducing these cues accurately
is a difficult technical problem. The cues that stem
from the complex interaction between sound waves and
the human body are particularly important but difficult
to reproduce.
Listeners experience sound in three dimensions: left/right,
up/down, and near/far (azimuth, elevation and range).
Typical two-speaker systems can control only the left/right
aspect. Even state-of-the-art "three-dimensional"
sound systems generally can only locate sounds on
a circle around a listener, and not in all three dimensions.
Among the challenges to creating true three-dimensional
sound fields is that each person's spatial sound cues
are strongly influenced by individual physical factors
such as the shape and position of their ears. These
factors -- which are captured by so-called Head-Related
Transfer Functions (HRTFs) -- vary greatly from person
to person. To mass-produce digitalsystems that accurately
reproduce three-dimensional sound fields requires
information about an individual listener's HRTF. The
new database provides the information that engineers
need for their designs.
Said Duda, "I believe that this customization of systems
to individual characteristics represents an important
and achievable goal for computer technology. Our current
NSF-supported work with colleagues at the University
of Maryland and Duke University is taking the next
step toward this goal by using computer vision techniques
and high-performance computing to obtain personalized
HRTFs."
To develop the database, Duda and Algazi meticulously
measured 45 different people to see exactly how the
sizes and shapes of their ears and bodies influenced
the sounds that reached their ears. Acoustic measurements
were stored in a database, together with measurements
of the size and shape of the listener's ears, heads
and torsos.
By knowing how a click pattern gets changed on the
way to a listener's ears, an engineer can modify any
sound presented over headphones to make it seem to
be coming from a particular location in space. Because
people have individual sizes and shapes, the modifications
must be individually tailored, much as eyeglasses
must be individually fit. Lacking data, engineers
previously have had to base their designs on an "average"
set of values, with results for the listeners similar
to using poorly fit eyeglasses. The new database will
provide the engineers with the information to properly
adjust their designs to account for individual differences.
The information is freely available for research or
commercial use on a compact disk or can be downloaded
from the Internet.
For more information, see http://interface.cipic.ucdavis.edu/.
Editors: The entire HRTF database can be downloaded
from: http://interface.cipic.ucdavis.edu/CIL_html/CIL_HRTF_database.htm.
|