Echosounding ,shallow seismic reflection and underwater sonographic investiga...
PosterICAD_4
1. Tracking moving sounds: perception of spatial figures
CIRMMT – MIL– McGill University CNRS – LaBRI – Université de Bordeaux(1) (2)
Cédric Camier (1)
– François-Xavier Féron (2)
– Catherine Guastavino (1)– David Romblom (1)
( ) Rotation table used by Stockhausen
for Kontakte (1958-1960)
( ) Excerpt of the 241 figures drawn by
Stockhausen for Cosmic pulses (2007)
( ) 48-loudspeaker configuration in the hemi-anechoic
SAL at CIRMMT
( ) Interface for collecting participant’s answers
2 3
4
Context and Objectives
Apparatus and Procedure Simulations and Stimuli
Results and Comments
Emergence of electroacoustic music in the 1950s: composers (e.g. Stochausen,
Xenakis, Nunes etc) became interested in positioning and moving sounds in
space (spatialized composition, spatial interpretation)
Composers have to render their pieces with various loudspeaker configurations,
which could lead to unexpected results. It has to be correctly anticipated
[A. Van de Gorne, “L’interprétation spatiale. essai de formalisation méthodologique.” Demeter, 2002.]
Immersion was reported from composers as one of the most desirable effect
produced by a spatial reproduction system which was linked to source width and
spatial reverberation simulation
[N. Peter, Sweet [re]production Ph.D. dissertation, McGill University, 2010.]
From recent previous works: neither observed effect of reverberation on
upper limit, nor on velocity discrimination
[C. Camier et al. , Does the reverberation affect upper limit for auditory motion perception, ICAD, 2015.]
[I. Frissen et al., “Auditory velocity discrimination in the horizontal plane at very high velocities,”
Hearing research, vol. 316, pp. 94–101, 2014.]
Aim: to determine under which conditions of spatial reproduction system
and of reverberation, simple spatial figures (such as circles,
squares and triangles) can be perceived by a listener positioned
in the center of a loudspeaker array
A 48-speaker circular array with a diameter
of 3.5 m in the hemi-anechoic Spatial Audio
Lab of the Centre for Interdisciplinary
Research in Music Media and Technology (CIRMMT)
Precomputed audio renderings with Matlab and played
by the Max/Msp interface
Task: to indicate which of three figures they perceived (circle, triangle
or square) using a 3-AFC (3 alternative forced choice)
5 variables were manipulated in a 4*12 min series of experiments:
Rotation direction (CW, CCW)
Loudspeaker sub-array (4, 16, 48 spk)
Reverberation (dry, reverberation simulation)
Spatial rendering techniques (VPAB, WFS)
Source velocity (0.5, 1, 2 rot/s)
The 13 participants were audio researchers or contemporary music and
electroacoustic composers
Early reflections Late reverberation (diffuse field)
Diffuse field model
Dynamic
mirror
images
Spatial renderer implemented in Matlab:
Modified VBAP (Angle-Based Amp. Panning)
with a delay & gain propagation computation
from the source locations to the speaker array
Sample-rate WFS computation with a spatial
window
Early reflections are computed from:
Dynamic mirror image sources
Coming from a virtual 4-wall configuration
From which each reflection coefficient is frequency-dependent
and different from the others
Diffuse field are computed from:
A 1st-order spherical impulse response measured from a concert
hall and treated to remove the direct path and early reflections.
Per-channel, frequency-dependent decorrelation filters based
on the statistical description of reverberation.
Array geometry is used in filter design for any N-loudspeaker
array
Stimuli source is a pass-band filtered white noise:
Cut-off frequencies have been selected in order to minimizing
differences in loudspeaker responses and acoustical paths
Factorial 2-way ANOVA reveal significative effects of:
Trajectory figures (F=4.08, p<0.018)
Speaker configuration (F=14.06, p~=0)
Interaction between these two factors (F=7.57, p~=0)
Velocity (F=23.05, p~=0)
Spatial renderer (F=23.81, p~=0)
Interaction between these two factors (F=3.96, p<0.02)
Interaction between participants and velocities and
between participants and trajectory figures
Comments from participants:
“Transition from a figure perception to a rhythm perception according to
the velocity”
“Harder in low velocities” vs “easier in low velocities”
“The reverberation seems to compensate for the holes to the benefit of
a coutinuous trajectory when the number of speakers is small”
1
0.5 rot/s 1 rot/s 2 rot/s
0.4
0.6
0.8
1
Velocity/Session
Session1
Session2
Session3
Session4
0.5 rot/s 1 rot/s 2 rot/s
0.4
0.6
0.8
1
Velocity/Spat renderer (for 48spk)
Vbap
WFS
0.5 rot/s 1 rot/s 2 rot/s
0.4
0.6
0.8
1
Correctanswerrate
Velocity/Rev
dry
Rev
0.5 rot/s 1 rot/s 2 rot/s
0.4
0.6
0.8
1
Velocity/Config.
48spk
16spk
4spk
Circ Triangle Square
0.4
0.6
0.8
1
Figure/Config. (for VBAP)
48spk
16spk
4spk
Correct answers rate over 13 participants classified
in various paired conditions. ( ) denotes significative
differences for a given x-axis condition
Main conclusions:
Simple figures (circles, triangles) are better perceived and
16 spk is good loudspeaker configuration
Confusions in the 4-spk config., especially at low velocity
WFS renderings facilitate the figure perception at low velovity
No observed effect of the modeled reverberation