From: Mart Lubbers <mart@martlubbers.net>
Date: Mon, 29 May 2017 14:57:39 +0000 (+0200)
Subject: elaborate on sections
X-Git-Url: https://git.martlubbers.net/?a=commitdiff_plain;h=0309ddbfdcaa0d8ccbeef78b5b06ff76894b173d;p=asr1617.git

elaborate on sections
---

diff --git a/conclusion.tex b/conclusion.tex
index 9d5176f..d1dd736 100644
--- a/conclusion.tex
+++ b/conclusion.tex
@@ -1,18 +1,33 @@
-\section{Conclusion}
+\section{Conclusion \& Future Research}
 This research shows that existing techniques for singing-voice detection
 designed for regular singing voices also work respectably on extreme singing
 styles like grunting. With a standard \gls{ANN} classifier using \gls{MFCC}
-features a performance of $85\%$ can be achieved. When applying smoothing this
-can be increased until\todo{results}.
+features a performance of $85\%$ can be achieved which is similar to the same
+techniques on regular singing. This means that it might be suitable as a
+pre-processing step for lyrics forced alignment.
+
+Future interesting research includes doing the actual forced alignment. This
+probably requires entirely different models. The models used for real speech
+are probably not suitable because the acoustic properties of a regular singing
+voice is very different from a growling voice, let alone speech.
+
+Secondly, it would be interesting if a model could be trained that could
+discriminate a singing voice for all styles of singing including growling.
+Moreover, it is possible to investigate the performance of detecting growling
+on regular singing-voice trained models and the other way around.
 
 %Discussion section
 \section{Discussion}
-Singing-voice detection can be seen as a crude way of
-genre-discrimination.\todo{finish}
+The dataset used is not very big. Only three albums are annotated and used
+as training data. The albums chosen do represent the ends of the spectrum and
+therefore the resulting model can be very general. However, it could also mean
+that the model is able to recognize three islands in the entire space of
+grunting. This does not seem the case since the results show that totally alien
+data also has a good performance.
 
-\todo{Novelty}
-\todo{Weaknesses}
-\todo{Dataset is not very varied but\ldots}
+The model clearly has trouble with pauses between singing.
 
-\todo{Doom metal}
-%Conclusion section
+\emph{Singing}-voice detection and \emph{singer}-voice Singing-voice detection
+can be seen as a crude way of genre-discrimination. Therefore it be
+generalizable to extensive genre recognition
+might.
diff --git a/methods.tex b/methods.tex
index 0a226a1..0e557df 100644
--- a/methods.tex
+++ b/methods.tex
@@ -211,6 +211,7 @@ batch size of $32$.
 
 \section{Results}
 \subsection{\emph{Singing} voice detection}
+
 \begin{table}[H]
 	\centering
 	\begin{tabular}{rccc}
@@ -234,6 +235,7 @@ batch size of $32$.
 \end{figure}
 
 \subsection{\emph{Singer} voice detection}
+
 \begin{table}[H]
 	\centering
 	\begin{tabular}{rccc}
@@ -249,3 +251,5 @@ batch size of $32$.
 	\end{tabular}
 	\caption{Multiclass classification results (accuracy (loss))}
 \end{table}
+
+\subsection{Alien data}