intro.tex

   1 \section{Introduction}
   2 The primary medium for music distribution is rapidly changing from physical
   3 media to digital media. The \gls{IFPI} stated that about $43\%$ of music
   4 revenue arises from digital distribution. Another $39\%$ arises from the
   5 physical sale and the remaining $16\%$ is made through performance and
   6 synchronisation revenues. The overtake of digital formats on physical formats
   7 took place somewhere in 2015. Moreover, ever since twenty years the music
   8 industry has seen significant growth
   9 again\footnote{\url{http://www.ifpi.org/facts-and-stats.php}}.
  10
  11 There has always been an interest in lyrics to music alignment to be used in
  12 for example karaoke. As early as in the late 1980s karaoke machines were
  13 available for consumers. While the lyrics for the track are almost always
  14 available, an alignment is not and it involves manual labour to create such an
  15 alignment.
  16
  17 A lot of this musical distribution goes via non-official channels such as
  18 YouTube\footnote{\url{https://youtube.com}} in which fans of the performers
  19 often accompany the music with synchronized lyrics. This means that there is an
  20 enormous treasure of lyrics-annotated music available but not within our reach
  21 since the subtitles are almost always hardcoded into the video stream and thus
  22 not directly usable as data. Because of this interest it is very useful to
  23 device automatic techniques for segmenting instrumental and vocal parts of a
  24 song, apply forced alignment or even lyrics recognition on the audio file.
  25
  26 These techniques are heavily researched and working systems have been created
  27 for segmenting audio and even forced alignment (e.g.\ LyricSynchronizer%
  28 \cite{fujihara_lyricsynchronizer:_2011}). However, these techniques are designed
  29 to detect a clean singing voice and have not been tested on so-called
  30 \emph{extended vocal techniques} such as grunting or growling. Growling is
  31 heavily used in extreme metal genres such as \gls{dm} but it must be noted that
  32 grunting is not a technique only used in extreme metal styles. Similar or equal
  33 techniques have been used in \emph{Beijing opera}, Japanese \emph{Noh} and but
  34 also more western styles like jazz singing by Louis
  35 Armstrong\cite{sakakibara_growl_2004}. It might even be traced back to viking
  36 times. For example, an arab merchant visiting a village in Denmark wrote in the
  37 tenth century\cite{friis_vikings_2004}:
  38
  39 \begin{displayquote}
  40         Never before I have heard uglier songs than those of the Vikings in
  41         Slesvig. The growling sound coming from their throats reminds me of dogs
  42         howling, only more untamed.
  43 \end{displayquote}
  44
  45 %Literature overview / related work
  46 \section{Related work}
  47 Applying speech related processing and classification techniques on music
  48 already started in the late 90s. Saunders et al.\ devised a technique to
  49 classify audio in the categories \emph{Music} and \emph{Speech}. They was found
  50 that music has different properties than speech. Music has more bandwidth,
  51 tonality and regularity. Multivariate Gaussian classifiers were used to
  52 discriminate the categories with an average performance of $90\%%
  53 $\cite{saunders_real-time_1996}.
  54
  55 Williams and Ellis were inspired by the aforementioned research and tried to
  56 separate the singing segments from the instrumental
  57 segments\cite{williams_speech/music_1999}. This was later verified by
  58 Berenzweig and Ellis\cite{berenzweig_locating_2001}. The latter became the de
  59 facto literature on singing voice detection. Both show that features derived
  60 from \gls{PPF} such as energy and distribution are highly effective in
  61 separating speech from non-speech signals such as music. The data used was
  62 already segmented.
  63
  64 Later, Berenzweig showed singing voice segments to be more useful for artist
  65 classification and used an \gls{ANN} (\gls{MLP}) using \gls{PLP} coefficients
  66 to detect a singing voice\cite{berenzweig_using_2002}. Nwe et al.\ showed that
  67 there is not much difference in accuracy when using different features founded
  68 in speech processing. They tested several features and found accuracies differ
  69 less that a few percent. Moreover, they found that others have tried to tackle
  70 the problem using myriads of different approaches such as using \gls{ZCR},
  71 \gls{MFCC} and \gls{LPCC} as features and \glspl{HMM} or \glspl{GMM} as
  72 classifiers\cite{nwe_singing_2004}.
  73
  74 Fujihara et al.\ took the idea to a next level by attempting to do \gls{FA} on
  75 music. Their approach is a three step approach. The first step is reducing the
  76 accompaniment levels, secondly the vocal segments are separated from the
  77 non-vocal segments using a simple two-state \gls{HMM}. The chain is concluded
  78 by applying \gls{Viterbi} alignment on the segregated signals with the lyrics.
  79 The system showed accuracy levels of $90\%$ on Japanese music%
  80 \cite{fujihara_automatic_2006}. Later they improved hereupon%
  81 \cite{fujihara_three_2008} and even made a ready to use karaoke application
  82 that can do the this online\cite{fujihara_lyricsynchronizer:_2011}.
  83
  84 Singing voice detection can also be seen as a binary genre recognition problem.
  85 Therefore the techniques used in that field might be of use.  Genre recognition
  86 has a long history that can be found in the survey by
  87 Sturm\cite{sturm_survey_2012}. It must be noted that of all the $485$ papers
  88 cited by Sturm only one master thesis is applying genre recognition on heavy
  89 metal genres\cite{tsatsishvili_automatic_2011}.
  90
  91 Singing voice detection has been tried on less conventional styles in the past.
  92 Dzhambazov et al.\ proposed to align long syllables in Beijing Opera to the
  93 audio\cite{dzhambazov_automatic_2016}. Beijing Opera sometimes contains
  94 growling like vocals. Dzhambazov also tried aligning lyrics to audio in
  95 classical Turkish music\cite{dzhambazov_automatic_2014}.
  96
  97 \section{Research question}
  98 It is debatable whether the aforementioned techniques work because the
  99 spectral properties of a growling voice is different from the spectral
 100 properties of a clean singing voice. It has been found that growling voices
 101 have less prominent peaks in the frequency representation and are closer to
 102 noise than clean singing\cite{kato_acoustic_2013}. This leads us to the
 103 research question:
 104
 105 \begin{center}\em%
 106         Are standard \gls{ANN} based techniques for singing voice detection
 107         suitable for non-standard musical genres like \gls{dm} and \gls{dom}?
 108 \end{center}