Content uploaded by Takeshi Naemura
Author content
All content in this area was uploaded by Takeshi Naemura on Jul 10, 2014
Content may be subject to copyright.
ISBN 88-7395-155-4 © 2006 ICMPC
1
Alma Mater Studiorum University of Bologna, August 22-26 2006
A Proposal on a Learning System to Realize
Maestro’s Favorite SPL Balance
Kumiko Morimura Makoto Iida Takeshi Naemura Hiroshi Harashima
Graduate School of
Interdisciplinary Information
Studies,
The University of Tokyo
Graduate School of
Mechanical Engineering,
The University of Tokyo
Graduate School of
Information Science and
Technology,
The University of Tokyo
Graduate School of
Interdisciplinary
Information Studies,
The University of Tokyo
kumikom@hc.ic.i.u-tokyo.ac.jp Tokyo, Japan Tokyo, Japan Tokyo, Japan
ABSTRACT
Choruses have been lead by maestros who are prominently
talented with musical skills. Their leading methods are clear
and the whole pictures of the music are always in the hands
of the conductors. It is difficult for ordinary people to know
the secret. We used information technology to build a
scientific learning system for chorus so that anybody can
practice by her/himself. Here we focused on sound
pressure level (SPL) balance in a chorus, and tried to
construct a learning system to realize maestro’s favorite
SPL balance. Six-channel choral voices were obtained
simultaneously, recorded in a computer, and presented to a
maestro so that he can adjust singers’ SPL balance to his
taste by using a digital mixer. An image of adjusted choral
voices was projected on the screen as a graphic, and the
singers sang watching the image and the graphic of their
actual SPL at a time. By using this system, singers were
able to improve their chorus skills scientifically.
Keywords
Chorus, multiple-channel recording, SPL balance, Maestro,
visualization
INTRODUCTION
Chorus trainings have been done empirically by few
transcendent maestros who have preeminent talent. It is a
special technique that could be done only by particular
persons, and the method hasn’t been revealed. Each singer
could practice pitch recognition or rhythm on her own, but
not the sound pressure level (SPL) balance because the
whole picture of the choral sound is in the maestro’s hands
and nobody else could touch it.
The authors want to make chorus practice more scientific
and something anybody can reach and try, we take an
interdisciplinary approach to research on chorus. One of
the reasons why chorus has not been studied scientifically is
difficulty in obtaining individual singing voice in a chorus
so that researchers could watch its behavior when singers try
to make harmonious sound.
In this study, we try to solve the problem by using
multi-channel recording system. We focus on SPL balance
among parts, and let a maestro adjust SPL balance of choral
voices using digital mixer, analyze the data and try to show
the result quantitatively.
Then, we think of a new method to visualize the sound. We
use graphics to show maestro’s favorite SPL balance to the
singers, and an image of each singer’s actual SPL is also
shown in real time on the same screen.
RELEVANT STUDIES
Relevant studies on choral music have been done mostly by
Sundberg, Ternstrom and his fellows [1]. They discussed
accuracy of musical intervals of chorus singers taking SPL
In: M. Baroni, A. R. Addessi, R. Caterina, M. Costa (2006) Proceedings
of the 9th International Conference on Music Perception & Cognition
(ICMPC9), Bologna/Italy, August 22-26 2006.©2006 The Society for
Music Perception & Cognition (SMPC) and European Society for the
Cognitive Sciences of Music (ESCOM). Copyright of the content of an
individual paper is held by the primary (first-named) author of that paper.
All rights reserved. No paper from this proceedings may be reproduced or
transmitted in any form or by any means, electronic or mechanical,
including photocopying, recording, or by any information retrieval
systems, without permission in writing from the paper's primary author.
No other part of this proceedings may be reproduced or transmitted in any
form or by any means, electronic or mechanical, including photocopying,
recording, or by any information retrieval system, without permission in
writing from SMPC and ESCOM.
ICMPC9 Proceedings
ISBN 88-7395-155-4 © 2006 ICMPC
1500
of reference sound into account. They measured
fluctuation of musical intervals to see how accurately
singers adjust their intervals when listening to the reference
sound at different levels. The result was that at the level of
25 dB of the reference sound, their pitch tolerance was
smallest. But, because they used the recorded sound as the
reference sound, one’s own voice will not be heard in the
reference sound, and there is no feedback from him/her to
the entire sound.
Ternstrom et al. have been doing their research on SOR (Self
to Others Ratio), which is the ratio of someone’s SPL toward
the others’ SPL of singing voices [2]. One cannot listen to
his own voice in the choir, because the recorded voices of
others are heard through the headphones when one sings.
As shown above, they used recorded voice as the reference
sound, and therefore, interaction between one singer and
other singers is not discussed. Only a few researches like
Jers’ have been done using multiple-channel recording
system to acquire individual voices in a chorus [3]. But as
far as I know, his research hasn’t dealt with interactive affect
of individual voices and the whole choir sound, especially
on SPL.
Singers in a choir sing listening to the other voices [4].
When we listen to the music played in an appropriate SPL,
we feel comfortable. When we listen to the chorus played
in an appropriate SPL balance, we feel at ease. Then, what
kind of SPL balance do we expect, when listening to a
chorus?
EXPERIMENTS AND RESULTS
I am going to tell the procedures and results of three experiments
we executed. The first experiment was done to acquire
multi-channel choral voices. The second experiment was to
acquire maestro’s favorite SPL balance. And the third one was to
try and evaluate the new learning system on the chorus training.
Experiment 1
In this paper, multiple-channel recording was done with the
cooperation of a 6 ladies chorus group for acquiring
multi-channel choral voices. A headset microphone was
attached on each singer, and each voice was recorded in
6-channel recording system as she sings in the chorus.
Figure 1 is the layout plan of the first experiment. Figure 1
indicates how singers stood in front of the conductor and the
arrangement of audio equipment. Singers stood in a line
from right to left (soprano, mezzo-soprano, and alt), and the
distance of the singers from the conductor was
approximately 2m. Audio equipment was set in front of the
singers. Singers wore the headset microphones. A
receiver of the headset microphone was fixed at an inch
distance from her mouth. The headset microphone has the
flat frequency property within 20 Hz to 20,000 Hz as tested
at the factory. In this experiment, all the microphones were
calibrated using a sound level meter.
Figure 1 Layout plan of the experiment 1.
Each singing voice was acquired by an individual headset
microphone (AKG C420), and after going through the
phantom power supply for stability, it goes through the
pre-amplifier (YAMAHA MLA7) and goes into the PC via
Fire Wire Audio Interface (MOTU 828mk2). A software
(AudioDesk) was used for recording, synthesizing, and
replaying. Their voices were recorded into the PC in 6
channels and analyzed.
Experiment 2
The recorded voices were presented to the maestro from the
computer to his headphones via the digital mixer. Using
this mixer, the maestro could adjust each singer’s SPL by
moving each lever that corresponds to a singer’s voice up
and down. The experimenter wrote down the position of
the gauges of the levers after the maestro set them at the
position he likes.
We had Maestro Kenji Otani, a well-known chorus
conductor in Japan, as the subject of this experiment.
Figure 2 is the picture of Maetro Otani sitting in front of the
mixer.
ICMPC9 Proceedings
ISBN 88-7395-155-4 © 2006 ICMPC
1501
Figure 2 Maestro Otani sits in front of the digital mixer to
adjust the SPL balance.
Maestro Otani sat in front of the digital mixer wearing
headphones on the head and moved the levers of the mixing
board up and down listening to the 6-channel voices
presented through the headphones. Sound presented
formed a chord named F-Dur (F sung by altos, A sung by
mezzo-sopranos, and C sung by sopranos), with two singers
singing in each part. He was allowed to listen to the sound
as long as he liked, in other words, he could use as much
time as he needed to realize his favorite SPL balance on the
mixer. Trials were done for 15 times with few minutes
brake after each trial. Data were taken each time after he
showed his favorite balance, collected, and analyzed later.
-25
-20
-15
-10
-5
0
5
10
Sop1 Sop2 M ez1 M ez2 Alto1 Alto2
channel
SPL(dB)
aou1
2
3
4
5
average
Figure 3 Maestro Otani’s favorite SPL balance
Figure 3 shows the tendency he showed of his favorite SPL
balance. X axis is channel of the singers and Y axis is
sound pressure level (dB). Each datum shows how he
adjusted each voice of the channel. The solid line is the
shape of the averaged SPL balance the maestro showed after
many times’ trials of adjusting.
Experiment 3
For the experiment 3, the authors tried to build a new system
to show the maestro’s favorite SPL balance and learn it.
We set two computers and two beam projectors to visualize
the images of the sound. From the computer (A), an image
of maestro’s favorite SPL balance was projected through the
beam projector (A) onto the screen. We also projected the
actual SPL of the singers’ voices in real time onto the same
screen through the beam projector (B) from the computer
(B). Singers sang watching both graphics and trying to
adjust their SPL to meet the maestro’s favorite SPL. Figure
4 is an image of the whole system with 2 computers, 2
projectors, and singers watching the screen.
Figure 4 An image of the whole system with two computers
and two beam projectors
Evaluation:
Evaluation test was done in the following procedures.
1. Singers were asked to sing as usual.
2. Singers watch the graphic image shown on the screen
while singing in order to set their SPL to meet the level
Maestro showed.
3. Singers sing at the level they learned using the system
by memory not watching the images.
We measured the SPL before and after they used the system.
By comparing both SPLs, we can tell if the system worked
effectively to improve their chorus skills.
Figure 5 is the result of the evaluation test and shows how
their skills were improved by using this system. X axis is
channel of the singer and Y axis is SPL (dB). The dotted
line is their SPL when they sang watching the image and the
solid line is their SPL when they sang by memory not
watching the images after practice.
75
80
85
90
95
100
105
Sop1 Sop2 Mez1 Mez2 Alto1 Alto2
chan nel
SPL (dB )
realizing
watching
ICMPC9 Proceedings
ISBN 88-7395-155-4 © 2006 ICMPC
1502
Figure 5 Singers SPL balance was very close to maestro’s
SPL balance after they practiced many times watching the
image.
CONCLUSION
The authors executed three experiments in order to propose
a new learning system to realize maestro’s favorite SPL
balance, so that the chorus singers could improve their skills
using it. We analyzed the data to see what kind of balance a
maestro would like to have, and visualized it in graphic.
Singers’ actual SPL were shown at the same time on the
same screen. We did a training lesson on a chorus group
using the system we proposed and the data showed the
system was effective to learn the maestro’s favorite SPL
balance and improve their chorus skills.
FUTURE WORK
The authors were able to make singers improve their chorus
skills, especially on SPL balance by using the system that we
propose. The system is still just germinated, and has much
to be improved. The system itself is too big to carry around
and too complicated for everyone to use. We would like to
make it handy and have better interface.
ACKNOWLEDGMENTS
I would deeply appreciate Harashima-Naemura Lab. to let
me use their facility to execute these experiments. I also
thank Maestro Otani and a chorus group “Ryf” to have let
me do my experiments on them.
REFERENCES
[1] Sundberg, J. (1989). The Science of the Singing Voice.
Northern Illinois University Press.
[2] Ternstrom. S., (1999). Preferred SOR in Choral Singing.
J Acoust Soc Am, June, 105(6), 3563-3574 (1999)
[3]Jers, Harald. (2004). Intonation Analysis of a
Multi-Channel Recording. Proceedings of Joint
Baltic-Nordic Acoustics Meeting 2004, 8-10 June (2004).
[4]Morimura, K., Iida, M., Naemura, T., & Harashima, H.
(2006). A Muti-channel Recording and Presentation System
to realize a good SPL Balance in a Chorus. Proceedings of
The 9th Western Pacific Acoustics Conference, June 26-28,
2006.
