Information Technology Reference
In-Depth Information
Finally we decided to apply a vision-based technique. The spatial (3-D)
position of a stick device is obtained by looking at the infrared LED marker
attached at its head. The height of the device over the table is estimated by
applying a triangulation method to the stereo images taken from two cam-
eras to which IR filters (infrared pass and visible light blocking filters) are
attached.
Meanwhile, the motion of the stick device is captured by an accelerometer
provided in Wii Remote. This means that it is needed for the system to have
correspondence between motion (sensor-based) data and position (vision-
based) data. However, images that are taken from cameras are in grayscale
and no additional features are available. There is no simple way to identify
which stick device it is. The system can easily make a misinterpretation when
multiple users participate in a session and multiple devices are overlapped
vertically over Sound Table.
In order to solve this problem, the system keeps comparing 3-D motion fea-
tures that are extracted from camera input images with those from the accel-
erometer, and decides one-to-one correspondences between LED marker
images and stick devices. Even if a correspondence is wrong at some point,
it can be corrected properly.
8.4.2 Positioning of Sounds
8.4.2.1 Algorithm
Sound Table is designed to be used by multiple users. Multiple different
sounds may appear at the same time, where each of the sounds is associated
with an individual user's action. It is essential for the system to help users
identify to whom the sound objects belong.
While 2-channel stereo and 5.1-channel surround sound frameworks for
creation of a virtual sound space are available and common in these days,
the best “spot” for listening is fixed. If the listener is out of the spot, a reality
of the sound space cannot be maintained any more.
Sound Table adopts 16 speakers to overcome this drawback (see Figure
8.2). Sounds can be positioned exactly where they should appear on the table.
The feel of sound presence is maintained wherever the user stands.
Here the distance between two adjacent speakers is 24 cm. Needless to
say, if a sound comes out just at the position where a speaker is placed,
it is not satisfactory. The system controls the loudness of each speaker so
they can feel the sound at any position irrelevant to the speaker positions.
We extendedly apply the well-known, sine-cosine pan law provided for
regular stereo sound management to this spatial (speaker array) sound
management.
The conventional sine-cosine pan law is explained as follows: For a pair of
two adjacent speakers, the loudness of one speaker is determined by multi-
plying the input by the sine of a control value, and that of the other speaker
