asr.tex

   1 %&asr
   2 \usepackage[toc,nonumberlist,acronyms]{glossaries}
   3 \makeglossaries%
   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 \newacronym{RELU}{ReLU}{Rectified Linear Unit}
  21 \newglossaryentry{dm}{name={Death Metal},
  22         description={is an extreme heavy metal music style with growling vocals and
  23         pounding drums}}
  24 \newglossaryentry{dom}{name={Doom Metal},
  25         description={is an extreme heavy metal music style with growling vocals and
  26         pounding drums played very slowly}}
  27 \newglossaryentry{FT}{name={Fourier Transform},
  28         description={is a technique of converting a time representation signal to a
  29         frequency representation}}
  30 \newglossaryentry{MS}{name={Mel-Scale},
  31         description={is a human ear inspired scale for spectral signals}}
  32 \newglossaryentry{Viterbi}{name={Viterbi},
  33         description={is a dynamic programming algorithm for finding the most likely
  34         sequence of hidden states in a \gls{HMM}}}
  35
  36 \begin{document}
  37 \frontmatter{}
  38
  39 \maketitleru[
  40         course={(Automatic) Speech Recognition},
  41         institute={Radboud University Nijmegen},
  42         authorstext={Author:},
  43         righttextheader={Supervisor:},
  44         righttext={Louis ten Bosch},
  45         pagenr=1]
  46 \listoftodos[Todo]
  47
  48 \tableofcontents
  49
  50 \mainmatter{}
  51 %Berenzweig and Ellis use acoustic classifiers from speech recognition as a
  52 %detector for singing lines.  They achive 80\% accuracy for forty 15 second
  53 %exerpts. They mention people that wrote signal features that discriminate
  54 %between speech and music. Neural net
  55 %\glspl{HMM}~\cite{berenzweig_locating_2001}.
  56 %
  57 %In 2014 Dzhambazov et al.\ applied state of the art segmentation methods to
  58 %polyphonic turkish music, this might be interesting to use for heavy metal.
  59 %They mention Fujihara (2011) to have a similar \gls{FA} system. This method uses
  60 %phone level segmentation, first 12 \gls{MFCC}s. They first do vocal/non-vocal
  61 %detection, then melody extraction, then alignment. They compare results with
  62 %Mesaros \& Virtanen, 2008~\cite{dzhambazov_automatic_2014}. Later they
  63 %specialize in long syllables in a capella. They use \glspl{DHMM} with
  64 %\glspl{GMM} and show that adding knowledge increases alignment (bejing opera
  65 %has long syllables)~\cite{dzhambazov_automatic_2016}.
  66 %
  67
  68
  69 %Introduction, leading to a clearly defined research question
  70 \chapter{Introduction}
  71 \input{intro.tex}
  72
  73 \chapter{Methods}
  74 \input{methods.tex}
  75
  76 \chapter{Conclusion \& Discussion}
  77 \input{conclusion.tex}
  78
  79 %(Appendices)
  80 \appendix
  81 \input{appendices.tex}
  82
  83 \newpage
  84 %Glossaries
  85 \glsaddall{}
  86 \begingroup
  87 \let\clearpage\relax
  88 \let\cleardoublepage\relax
  89 \printglossaries{}
  90 \endgroup
  91
  92 \bibliographystyle{ieeetr}
  93 \bibliography{asr}
  94 \end{document}