Information Technology Reference
In-Depth Information
includes audio capturing techniques, the represen-
tation or encoding of the signal as a soundfield
(referred to as 'B-format'), and the decoding of
the signal during reproduction. Particular mi-
crophone patterns and positions, or a specially
built microphone with four specifically arranged
capsules (Soundfield Microphone), are required
to capture the signal in a compatible way. The
result is 4 signals conventionally referred to as
X, Z, Y, W in first order Ambisonics:
significant amount of computation if its trajectory
is to be smooth and uninterrupted. A balance needs
to be struck and compromises are necessary. It is
broadly accepted that update rates of 60 Hz and
a total delay time of up to 50 msec are acceptable
for acoustic virtual sound (Vorländer, 2008).
In many instances, a perceptual evaluation
referred to as the Just-Noticeable-Difference
(JND) is a useful instrument. This psychoacoustic
evaluation procedure can be used for pitch dif-
ferentiation or temporal differentiation, to name
just a few. In terms of spatial sound, it is useful to
know the accuracy of the human hearing system in
differentiating between the same sound-source at
different degrees in space. With this information,
previous calculations that were not perceptually
detected by the listener can now be disregarded.
Some useful JND values in this regard allows for
the reduction of redundant data, and resources
can be used for other processes such as sound
propagation. The performance of human listen-
ers in point-to-point localization on the azimuth
is most accurate in the frontal direction at 1º. On
the left/right axis, the performance diminishes
significantly to 10º. The rear direction is 5º and
the JND on the vertical plane is the least accurate
at 20º (Vorländer, 2008). Therefore, a computer
game or VR system can effectively render spatial
sound with diminished precision at certain loca-
tions around the listener.
Real-time binaural synthesis (the generation
of spatial sound for headphone reproduction),
is also a very resource demanding process, and
interesting techniques are used to address the
problem. One of the methods used is to preprocess
binaural sound for reproduction when the listener
reaches particular coordinates and orientation in
the virtual room. A best-matched filter is applied
as the listener's position changes. The key to this
approach is to make the changeover from filter to
filter as inaudible as possible. Fast convolution
is required no matter which real-time binaural
synthesis approach is taken. With many channels
being processed in parallel, and with the continu-
• X = Front minus Back
• Z = Up minus Down
• Y = Left minus Right
• W = A non-directional reference signal. Front
+ Up + Left + Back + Down + Right.
Although considered one of the most advanced
and realistic spatial reproduction systems avail-
able, the Ambisonics format has suffered from
commercial setbacks. These can be attributed
to bad timing in entering the marketplace, mis-
leading associations with ill-fated quadraphonic
techniques, and the lack of uptake by key music
industry players during its development. However,
recently the computer game industry and virtual
reality researchers have stimulated renewed inter-
est in Ambisonics.
real-time Processing
The real-time simulation of physical environments
in computer games and VR systems requires a
significant degree of real-time processing. Because
of the complexity of this task, a certain amount
of latency can be expected but limits must be
placed on the extent of the latency so as not to
thwart the user-experience. With current hardware,
latency issues remain a concern, especially when
attempting to reproduce a real-world scenario
that requires rapid and consistent refresh rates of
visual, aural, haptic, and motional events. Even
within a sound-only game, the continuous motion
of a sound source around the listener requires a
Search WWH ::




Custom Search