2 \usepackage[toc,nonumberlist,acronyms
]{glossaries
}
4 \newacronym{ANN
}{ANN
}{Artificial Neural Network
}
5 \newacronym{DCT
}{DCT
}{Discrete Cosine Transform
}
6 \newacronym{DHMM
}{DHMM
}{Duration-explicit
\acrlong{HMM
}}
7 \newacronym{FA
}{FA
}{Forced alignment
}
8 \newacronym{GMM
}{GMM
}{Gaussian Mixture Models
}
9 \newacronym{HMM
}{HMM
}{Hidden Markov Model
}
10 \newacronym{HTK
}{HTK
}{\acrlong{HMM
} Toolkit
}
11 \newacronym{IFPI
}{IFPI
}{International Federation of the Phonographic Industry
}
12 \newacronym{LPCC
}{LPCC
}{\acrlong{LPC
} derivec cepstrum
}
13 \newacronym{LPC
}{LPC
}{Linear Prediction Coefficients
}
14 \newacronym{MFCC
}{MFCC
}{\acrlong{MFC
} coefficient
}
15 \newacronym{MFC
}{MFC
}{Mel-frequency cepstrum
}
16 \newacronym{MLP
}{MLP
}{Multi-layer Perceptron
}
17 \newacronym{PLP
}{PLP
}{Perceptual Linear Prediction
}
18 \newacronym{PPF
}{PPF
}{Posterior Probability Features
}
19 \newacronym{ZCR
}{ZCR
}{Zero-crossing Rate
}
20 \newglossaryentry{dm
}{name=
{Death Metal
},
21 description=
{is an extreme heavy metal music style with growling vocals and
23 \newglossaryentry{dom
}{name=
{Doom Metal
},
24 description=
{is an extreme heavy metal music style with growling vocals and
25 pounding drums played very slowly
}}
26 \newglossaryentry{FT
}{name=
{Fourier Transform
},
27 description=
{is a technique of converting a time representation signal to a
28 frequency representation
}}
29 \newglossaryentry{MS
}{name=
{Mel-Scale
},
30 description=
{is a human ear inspired scale for spectral signals.
}}
31 \newglossaryentry{Viterbi
}{name=
{Viterbi
},
32 description=
{is a dynamic programming algorithm for finding the most likely
33 sequence of hidden states in a
\gls{HMM
}}}
39 course=
{(Automatic) Speech Recognition
},
40 institute=
{Radboud University Nijmegen
},
41 authorstext=
{Author:
},
42 righttextheader=
{Supervisor:
},
43 righttext=
{Louis ten Bosch
},
50 %Berenzweig and Ellis use acoustic classifiers from speech recognition as a
51 %detector for singing lines. They achive 80\% accuracy for forty 15 second
52 %exerpts. They mention people that wrote signal features that discriminate
53 %between speech and music. Neural net
54 %\glspl{HMM}~\cite{berenzweig_locating_2001}.
56 %In 2014 Dzhambazov et al.\ applied state of the art segmentation methods to
57 %polyphonic turkish music, this might be interesting to use for heavy metal.
58 %They mention Fujihara (2011) to have a similar \gls{FA} system. This method uses
59 %phone level segmentation, first 12 \gls{MFCC}s. They first do vocal/non-vocal
60 %detection, then melody extraction, then alignment. They compare results with
61 %Mesaros \& Virtanen, 2008~\cite{dzhambazov_automatic_2014}. Later they
62 %specialize in long syllables in a capella. They use \glspl{DHMM} with
63 %\glspl{GMM} and show that adding knowledge increases alignment (bejing opera
64 %has long syllables)~\cite{dzhambazov_automatic_2016}.
68 %Introduction, leading to a clearly defined research question
69 \chapter{Introduction
}
75 \chapter{Conclusion \& Discussion
}
76 \input{conclusion.tex
}
80 \input{appendices.tex
}
87 \let\cleardoublepage\relax
91 \bibliographystyle{ieeetr
}