Towards Home-Made Dictionaries for Musical Feature Extraction

Research output: Working paperResearch

116 Downloads (Pure)

Abstract

The majority of musical feature extraction applications are based on the Fourier transform in various disguises. This is despite the fact that this transform is subject to a series of restrictions, which admittedly ease the computation and interpretation of transform coefficients, but also imposes arguably unnecessary limitations on the ability of the transform to extract and identify features. However, replacing the nicely structured dictionary of the Fourier transform (or indeed other nice transform such as the wavelet transform) with a home-made dictionary is a dangerous task, since even the most basic properties are easily lost.
Original languageDanish
Publication statusPublished - 2003

Cite this

@techreport{2b407eb09c2d11db8ed6000ea68e967b,
title = "Towards Home-Made Dictionaries for Musical Feature Extraction",
abstract = "The majority of musical feature extraction applications are based on the Fourier transform in various disguises. This is despite the fact that this transform is subject to a series of restrictions, which admittedly ease the computation and interpretation of transform coefficients, but also imposes arguably unnecessary limitations on the ability of the transform to extract and identify features. However, replacing the nicely structured dictionary of the Fourier transform (or indeed other nice transform such as the wavelet transform) with a home-made dictionary is a dangerous task, since even the most basic properties are easily lost.",
author = "Harbo, {Anders La-Cour}",
year = "2003",
language = "Dansk",
type = "WorkingPaper",

}

TY - UNPB

T1 - Towards Home-Made Dictionaries for Musical Feature Extraction

AU - Harbo, Anders La-Cour

PY - 2003

Y1 - 2003

N2 - The majority of musical feature extraction applications are based on the Fourier transform in various disguises. This is despite the fact that this transform is subject to a series of restrictions, which admittedly ease the computation and interpretation of transform coefficients, but also imposes arguably unnecessary limitations on the ability of the transform to extract and identify features. However, replacing the nicely structured dictionary of the Fourier transform (or indeed other nice transform such as the wavelet transform) with a home-made dictionary is a dangerous task, since even the most basic properties are easily lost.

AB - The majority of musical feature extraction applications are based on the Fourier transform in various disguises. This is despite the fact that this transform is subject to a series of restrictions, which admittedly ease the computation and interpretation of transform coefficients, but also imposes arguably unnecessary limitations on the ability of the transform to extract and identify features. However, replacing the nicely structured dictionary of the Fourier transform (or indeed other nice transform such as the wavelet transform) with a home-made dictionary is a dangerous task, since even the most basic properties are easily lost.

M3 - Working paper

BT - Towards Home-Made Dictionaries for Musical Feature Extraction

ER -