~ICMC 2015 - Sept. 25 - Oct. 1, 2015 - CEMI, University of North Texas
Fuzzy Equalization of Musical Genres
Marie Gonzalez
Department of
Electrical Engineering
School of Engineering
Pontificia Universidad Cat6lica
de Chile
[email protected]
Patricio de la Cuadra
Center for Research in
Audio Technologies
Music Institute
Pontificia Universidad Cat6lica
de Chile
[email protected]
Rodrigo F. Cadiz
Center for Research in
Audio Technologies
Music Institute
Pontificia Universidad Cat6lica
de Chile
[email protected]
ABSTRACT
In this article, we propose an audio equalizer of musical genres based on fuzzy logic. Widely used audio playback software, such as VLC or iTunes, propose genre-specific equalization curves to be applied for the whole duration of the music. These curves are the same for all songs belonging to a
specific genre, and they do not take into account the specifics
and diversity of each song. We propose a different strategy.
Research in music information retrieval has revealed a significant number of audio descriptors that allow for the recognition and description of diverse musical genres. We use some
of these descriptors to feed a fuzzy logic inference system,
whose outputs are the required equalization levels for each
frequency band. The rules of the system are derived from the
analysis of a well known music database encompassing ten
different musical genres. Our results indicate that our approach works for songs that exhibit multiple genre characteristics, that are difficult to classify into one category, or that
mix genres.
1. INTRODUCTION
Audio equalization using digital signal processors has been
a subject of research for several decades. An equalizer aims
for the correction of the magnitude and the phase response of
an audio chain [1]. The equalization of audio is required in
a great number of distinct situations such as radio broadcasting, recording studios, or music listening while driving or at
home.
Equalization is useful to correct for uneven frequency responses of sound reproduction systems, which essentially means
static equalization curves for all genres. Consequently, it appears that equalization based on genre is unnecessary and not
even desirable. This argument strengthens considering all the
work and effort that mastering engineers have put into balancing the audio spectrum in the studio. However, we would
Copyright: 02015 Marie Gonzilez et al. This is an open-access article
distributed under the terms of the Creative Commons Attribution License 3.0
Unported, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original author and source are credited.
like to emphasize that genre-based equalization is a common
reality. Car radios and home audio playback systems contain
this kind of equalization. Very popular audio software such as
VLC or iTunes base their equalization curves on genre information, as it is shown in figure 1. We simply want to propose
a different strategy for genre equalization and we don't have
the intention at this moment of arguing against or in favor of
it.
These curves, although somewhat similar in different software, appear to be based mostly on experience and customized
by each vendor, and we have not found scientific or widely
accepted methodologies for their design. Similar curves are
also present in the majority of audio reproduction equipments
and car radios. Each one of these curves aims for a specific
musical genre such as rock, pop, jazz or electronic. The user
is supposed to select one of them when listening to a particular piece of music. However, what curves should be used
for music that mix several styles, songs that contain a significant amount of musical diversity, or when the genre one
is listening to does not appear in the lists of available equalizations? In order to tackle this problem, we propose a fuzzy
logic-based equalization scheme that allows for a more meaningful equalization of music, including mixed musical genres
or genres outside the realm of traditional music.
This article is structured as follows. First, fuzzy logic is presented as discussed as a valid technique for non-linear mappings of data. Next, we present our fuzzy logic-based audio equalizer including previous analysis of the audio data,
determination of fuzzy inference rules, output variables and
computational implementation. In the following section, we
describe one simple experiment we conducted in order to test
our equalization scheme and later the main results are presented. Finally, conclusions and future work are addressed.
2. FUZZY LOGIC
Fuzzy logic [2] [3] [4] [5] [6] is a concept derived from the
mathematical branch of fuzzy sets [7] that applies multi-valued
logic to sets or groups of objects. In a narrow sense, fuzzy
logic refers to a logical system than generalizes traditional
two-valued logic for reasoning under uncertainty, allowing
multiple values of truth. In a broad sense, it refers to all the
-134 -
