ï~~Proceedings of the International Computer Music Conference 2011, University of Huddersfield, UK, 31 July - 5 August 2011
TOWARDS A COMPREHENSIVE
FRAMEWORK FOR
ELECTRO-ACOUSTIC MUSIC ANALYSIS
Tae Hong Park, David Hyman, Peter Leonard, Phillip Hermans
Tulane University
Music Department
ABSTRACT
Computers have become instrumental in creating electroacoustic music. However, they have not yet significantly
augmented our ability to gain insight into the music
itself. This is not surprising as music analysis in general
is greatly facilitated by the availability of a "score," and
the majority of electro-acoustic music pieces, many of
which are timbre-centric, in fact have no score.
In response to this situation we have been motivated
to undertake research with the purpose of gaining a
greater understanding of timbre in order to aid in the
analysis of electro-acoustic music. We have taken a
twofold approach-exploitation of computer-based
music information retrieval (MIR) techniques and the
development of a comprehensive methodology for
analysis. The software portion consists of a "click-andgo" GUI called the EASY (Electro-Acoustic muSic
analYsis) Toolbox. The methodological component is
called SQEMA (Systematic and Quantitative Electroacoustic Music Analysis).
In this paper we will focus on updates to EASY and
SQEMA and discuss our goal of integrating the two. We
will also present new research into a technique we call
companded listening and a feature vector called sensory
dissonance analysis. Finally, we will present new results
of studies using SQEMA and EASY in analyzing classic
electro-acoustic works.
1. INTRODUCTION
The course of our research over the last several years has
been guided by two important ideas. The first is the
development of the EASY (Electro-Acoustic music
AnalYsis) Toolbox, a software system which employs
MIR techniques to assist in electro-acoustic music
analysis; the second, SQEMA (Systematic and
Quantitative Electro-Acoustic Music Analysis), is a
methodology for the analysis of electro-acoustic music.
These two tools were developed separately and can be
used independent of one another. With further research
and development, however, they have essentially
merged. As their interconnection has become clearer, we
have undertaken steps to integrate them into a unified
tool with the aim of developing a comprehensive
framework for electro-acoustic music analysis.
Ultimately, this framework will rely on software to
perform arduous tasks including formal segmentation,
clustering, and detection of events. In its final form we
foresee the system outputting a navigable map based on
strictly quantitative analysis, analogous to the pitch and
time based scores of traditional music. The organization
of copious data in this map form will allow the user to
focus on the subjective aspects of analysis, resulting in
more informed and stronger aesthetic interpretations.
This paper presents updates and refinements which
have arisen from further testing and development of
EASY and SQEMA. First we will give brief descriptions
of EASY and SQEMA and discuss the development of
the relationship between the two tools. Next we will give
an overview of new methods and feature vectors which
we have tested and implemented in EASY and SQEMA.
We will then present results of testing the system on
several pieces we have analyzed since our earlier studies.
Finally we will discuss our conclusions and ideas for
future work in this area.
2. EASY TOOLBOX AND SQEMA
2.1. Brief Overview of EASY and SQEMA
The EASY Toolbox is currently implemented in
MATLAB and features a GUI based "click-and-go"
interface. EASY can handle very large files, limited only
by available hard disk space, and includes transport
functions for audio playback. The output of
time/frequency domain analysis algorithms are displayed
via the following visualization formats: waveform, 3D
spectrogram, feature vector plotting,
clustering/segmentation, and the timbregram [6].
Building in part on the information produced by the
EASY Toolbox, we began developing the SQEMA
methodology. SQEMA stands for Systematic and
Quantitative Electro-Acoustic Music Analysis. SQEMA
is based on two complementary techniques: (a)
exploitation of MIR and (b) segmentation and reduction
of the complexity of a piece of music into simpler,
smaller, and more manageable parts. The EASY
Toolbox addresses technique (a), outputting quantitative
information that the user can employ to approach (b).
SQEMA employs the following stepwise model for
analysis: (I) Multiple Listenings, (II) High-Level
Analysis, (III) Mid-Level Analysis, (IV) Low-Level
Analysis, (V) Reexamination, (VI) Aesthetic
Interpretation. Following multiple listenings, we utilize
data from quantitative analysis to segment the piece into
166
0
