179 lines
8.2 KiB
TeX
179 lines
8.2 KiB
TeX
|
|
\documentclass{scrartcl}
|
|
\usepackage{enumitem}
|
|
\usepackage[british]{babel}
|
|
\usepackage[style=apa, backend=biber]{biblatex}
|
|
\DeclareLanguageMapping{british}{british-apa}
|
|
\usepackage{url}
|
|
\usepackage{float}
|
|
\restylefloat{table}
|
|
\usepackage{perpage}
|
|
\MakePerPage{footnote}
|
|
\usepackage{abstract}
|
|
\usepackage{graphicx}
|
|
% Create hyperlinks in bibliography
|
|
\usepackage{hyperref}
|
|
|
|
\usepackage[T1]{fontenc}
|
|
\usepackage[utf8]{inputenc}
|
|
\usepackage{blindtext}
|
|
\setkomafont{disposition}{\normalfont\bfseries}
|
|
|
|
|
|
\graphicspath{
|
|
{./resources/},
|
|
}
|
|
\addbibresource{~/PerryPerrySource/LaTeX/FYP_Bibliography.bib}
|
|
|
|
\newsavebox{\abstractbox}
|
|
\renewenvironment{abstract}
|
|
{\begin{lrbox}{0}\begin{minipage}{\textwidth}
|
|
\begin{center}\normalfont\sectfont\abstractname\end{center}\quotation}
|
|
{\endquotation\end{minipage}\end{lrbox}%
|
|
\global\setbox\abstractbox=\box0 }
|
|
|
|
\usepackage{etoolbox}
|
|
\makeatletter
|
|
\expandafter\patchcmd\csname\string\maketitle\endcsname
|
|
{\vskip\z@\@plus3fill}
|
|
{\vskip\z@\@plus2fill\box\abstractbox\vskip\z@\@plus1fill}
|
|
{}{}
|
|
\makeatother
|
|
|
|
\DeclareCiteCommand{\citeyearpar}
|
|
{}
|
|
{\mkbibparens{\bibhyperref{\printdate}}}
|
|
{\multicitedelim}
|
|
{}
|
|
|
|
\begin{document}
|
|
\title{Independant Project\\Descriptor Driven Concatenative Synthesis Tool}
|
|
\subtitle{\LARGE{Results Analysis}}
|
|
\author{Sam Perry\\U1265119}
|
|
\date{}
|
|
|
|
\maketitle
|
|
|
|
\section{Overview}
|
|
This document aims to provide reasoning for the sonic characteristics of
|
|
the audio produced in this project. Samples have been provided to
|
|
demonstrate the breadth of possibilities for the concatenator script.
|
|
Samples use a range of databases of varying sizes and aim to show both the
|
|
strengths of this script and areas that could be developed further to
|
|
improve results.
|
|
|
|
\section{Sample Analysis}
|
|
Samples are organized in folder named by their source\char`_target databases.
|
|
|
|
\subsection{Guitar\char`_Guitar}
|
|
Source Database: University of Iowa Sample Database - Guitar\\
|
|
Target Database: David Chaplin's Guitar Samples\\
|
|
|
|
These samples were generated with the intention of matching the clean
|
|
guitar to a clean guitar database. It was thought that in an ideal
|
|
situation this would produce a result similar to the original but with the
|
|
tonal qualities of the sampled guitar. The result is far from this as the
|
|
current descriptors do not provide a description that is sufficiently
|
|
accurate. However, The General contour of the clean guitar can be heard and
|
|
the ``plucky'' qualities of the sampled database are definitely present.
|
|
|
|
\subsection{Guitar\char`_Violin}
|
|
Source Database: University of Iowa Sample Database - Violin\\
|
|
Target Database: David Chaplin's Guitar Samples\\
|
|
Overall, result produced were smooth and for the majority of the samples
|
|
used, are similar to that of a stringed instrument. The occasional glitch
|
|
detracts from this and there are a number of reason for these small glitches.
|
|
At around 3 seconds, a high pitched screech can be heard that does not fit
|
|
in with it's neighbouring samples. When comparing this to the event in the
|
|
original file, it can be seen that this is a sharp decay and pause that is
|
|
analysed as an inharmonic signal with very different characteristics to
|
|
it's neighbouring samples. Use of a viterbi algorithm would most likely
|
|
help to minimise these types of error, alongside better handling of
|
|
inharmonic samples.
|
|
|
|
\subsection{Guitar\char`_Xylophone}
|
|
Source Database: University of Iowa Sample Database - Xylophone\\
|
|
Target Database: David Chaplin's Guitar Samples\\
|
|
This example demonstrated interesting timbral qualities due to the rich
|
|
variety of harmonic percussive sounds in the Xylophone database. By
|
|
matching primarily on pitch and spectral content, a mixture of transient
|
|
grains are concatenated with reasonable results.
|
|
|
|
\subsection{Multiple\char`_Vocal}
|
|
Source Database: Alex Harker Vocal Database\\
|
|
Target Database: David Chaplin's Guitar Samples, Rayman Spoken Word
|
|
Database\\
|
|
This database provided interesting results in terms of timbre as the outcome
|
|
is a mixture of vocal techniques. This sounds entirely synthetic which is
|
|
expected, and there are glitches as with all the results. However, the results
|
|
appear to be perceptually related to both the source and target databases.
|
|
Vocal qualities can be heard through the clicks, hums and other elements
|
|
that are combined. Continuity between grains is a significant problem as
|
|
selected grains do not flow in the same way that natural vocals might.
|
|
Again, comparing matches in a post matching process such as a viterbi
|
|
decoding stage would help to counter this effect.
|
|
A stretched version of Clean\char`_1 was included to demonstrate the results of
|
|
setting different values for matching and synthesis overlaps. By setting the
|
|
matching overlap to 8 and the synthesis overlap to 4, the synthesis was
|
|
stretched by a factor of 2.
|
|
|
|
\subsection{Spoken\char`_Xylophone}
|
|
Source Database: University of Iowa Sample Database - Xylophone\\
|
|
Target Database: Rayman Spoken Word Database\\
|
|
|
|
This example was generated during development and has been included as it
|
|
demonstrated an effective mix between the qualities of the two databases.
|
|
The pitch and envelope of the vocals can be heard whilst the timbre of the
|
|
Xylophone database is also clear. Glitchy pitch shifted grains can be heard
|
|
in the background as this was a problem with the project at the time and
|
|
has since been fixed by altering handling of inharmonic grains in the
|
|
picth shifter.
|
|
|
|
\subsection{Vocal\char`_Marimba}
|
|
Source Database: University of Iowa Sample Database - Xylophone\\
|
|
Target Database: Danny Darko Acapella Vocals Sample\\
|
|
This sample highlights a number of problems with the current implementation
|
|
of this algorithm.
|
|
There is a considerable amount of noise as a result of the intensity
|
|
scaling. This has increased quite sounds with low signal/noise ratios and
|
|
as a result this can be heard clearly in the results.
|
|
The clicks produced through poor performance of the pitch shifter have also
|
|
degraded the signal during synthesis. Replacing the pitch shifter with a
|
|
more effective algorithm would most likely improve results.
|
|
|
|
\subsection{Vocal\char`_Wind}
|
|
Source Database: University of Iowa Sample Database - Xylophone\\
|
|
Target Database: Danny Darko Acapella Vocals Sample\\
|
|
These examples are significantly less satisfying. Inharmonic noise from the
|
|
player taking breaths mixes poorly to generated a layer of breathy noise.
|
|
This is further emphasized by the pitch shifting of this noise. This
|
|
demonstrates an inherent problem with pitch shifting technique used to
|
|
transform grains, in that inharmonic audio is also transformed. This would
|
|
be greatly improved with the use of a better pitch shifting algorithm.
|
|
|
|
\section{Overall Performance}
|
|
Overall, it was difficult to generate results of a high quality due to the
|
|
current issues with the algorithm used. During development, much time was
|
|
spent increasing the speed of the algorithm to make it possible to create a
|
|
high enough quantity of results. On reflection, spending more time on
|
|
improving output quality may have yielded better results. However,
|
|
overall it is clear that this style of synthesis shows real potential for
|
|
generating a wide range of interesting sound and as a prototype or proof of
|
|
concept, the script achieves a reasonable quality of result.
|
|
|
|
\section{Database Sources}
|
|
\begin{table}[H]
|
|
\centering
|
|
\label{my-label}
|
|
\begin{tabular}{ll}
|
|
\textbf{Database Name} & \textbf{Accessed from} \\
|
|
University of Iowa Sample Database & http://theremin.music.uiowa.edu/MIS.html \\
|
|
Danny Darko Acapella Vocals Sample & http://www.dannydarko.net/category/remix-stems/\\
|
|
Alex Harker's Vocal Sample Database & Not Publicly Available \\
|
|
David Chaplin's Guitar Samples & Not Publicly Available \\
|
|
Rayman Spoken Word Database & Not Publicly Available
|
|
\end{tabular}
|
|
\end{table}
|
|
|
|
\end{document}
|