Please use headphones to listen to the binaural material presented here. The recordings have been slightly pre-equalized to ensured a natural reproduction on standard (i.e. diffuse field compensated) studio headphones.
Motion Tracked Binaural Sound (MTB), is a novel
head-related multi-channel microphone recording method proposed by
Algazi et al. in 2004 (J. Audio Eng. Soc., Vol. 52, No. 11, p.
1142-1156). It uses a number of microphones which are mounted on the
horizontal circumference of a (mostly) spherical rigid shell. This
shell leads to acoustical shadowing and temporal offsets, which are
encoded into the microphone signals. Hence, MTB recordings will
include some ‘binaural’ signal features such as interaural time
and level differences.
Further, the rotational symmetry of the MTB array allows for a dynamic sound reproduction: Therefore, the signals of two microphones which are located on opposite sides of the shell are presented via headphones to a listener. Then, horizontal head movements of the listener are tracked in real-time. If the listener turns his head, the current signal is exchanged against that of two other microphones which are closest to the new position of the listener’s ears (see picture).
- © Copyright??
If this exchange is done continuously and in
real-time, as it is possible with current methods of signal
processing, an almost natural listening experience - as with
conventional dynamic binaural reproduction  – can be achieved:
Sound sources are perceived outside of the head and at fixed positions
while moving one’s head. Depending on the number of microphones and
type of crossfading algorithm used for interpolating towards new head
orientations the process can result in very different impressions,
MTB’s reproduction quality is limited by the lack of the human external ears and by the fixed diameter of the spherical shell. Thus, one often observes problems with the perception of vertical direction and with the stability of sound source location. However, currently MTB is the only method for recording real acoustic scenes allowing for post-hoc individualized dynamic binaural cues. Additionally, MTB signals (live or recorded) may – at the same time – be rendered independently for an arbitrary number of listeners. Thus, it might be promising to apply MTB in various fields as e.g., documentation of sound installations, soundscape research, as new interactive and immersive (streaming) media format for live broadcasting or for marine acoustic localization.
The three videos present stimuli taken from an empirical systematic study which, for the first time, assessed the perceptual impact of the number of microphones, interpolation method and the contents used with the MTB method. The videos show the MTB player, a dedicated MTB rendering software developed at our department. The animation in the lower right shows the current orientation of the listener's head which continuously changed its orientation when the videos were created. In the upper part of the video, the results of the aforementioned study are shown. The variable highlighting is intended to make it easier to gain a quick overview of the most significant found effects. Since these effects were depending on the content, the video exists in three versions (for a source with pink noise, four instruments of a string quartet, and for a speaker (yet to come)). First, the differences of five crossfade algorithms are demonstrated for a constantly high number of microphones (24) and ordered for increasing quality. After that, the microphone effect is demonstrated using the most obvious one of the crossfade algorithms. Finally, the perceived best crossfade algorithm is presented, for which even an effect of the number of microphones could hardly be found.