Onset detection for drum loop classification

The main objective of this report is to be able to perform drum beat detection through a semi-automatic way to produce the drum samples and then use these drum samples to produce a general classifier that is able to identify the drum beats in any music track. An algorithm was first introduced to per...

Full description

Saved in:
Bibliographic Details
Main Author: Toh, Jackson Yit Chuan.
Other Authors: Yap Kim Hui
Format: Final Year Project
Language:English
Published: 2009
Subjects:
Online Access:http://hdl.handle.net/10356/17928
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-17928
record_format dspace
spelling sg-ntu-dr.10356-179282023-07-07T16:50:18Z Onset detection for drum loop classification Toh, Jackson Yit Chuan. Yap Kim Hui School of Electrical and Electronic Engineering Zhu Yongwei DRNTU::Engineering The main objective of this report is to be able to perform drum beat detection through a semi-automatic way to produce the drum samples and then use these drum samples to produce a general classifier that is able to identify the drum beats in any music track. An algorithm was first introduced to perform onset detection on percussive sounds for a small period of music track of about six seconds. The user is involved in labeling which of these percussive sounds are the bass and snare drum, hence a semi-automatic way approach in drum beat detection. A small database of 21 music tracks was used in this project for performance evaluation. Performance evaluation of the onset detection algorithm is done by evaluating the average precision rate and recall rate of the algorithm for all the music tracks in the database. The percussive onset detection algorithm had an average precision rate of 0.714 and average recall rate of 0.953. After the onset detection, the next task was to train a classifier using the previous drum samples detected previously by the algorithm. There were two different scenarios in achieving this task. The first scenario was to classify the target vector into 2 classes with bass and snare drum, and the second scenario was to classify the target vector into 3 classes with the bass drum, snare drum and other onsets which are neither bass nor snare. The results of these two scenarios and possibilities for future work to solve this issue will be discussed in this report. Bachelor of Engineering 2009-06-18T01:59:24Z 2009-06-18T01:59:24Z 2009 2009 Final Year Project (FYP) http://hdl.handle.net/10356/17928 en Nanyang Technological University 53 p. application/pdf
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic DRNTU::Engineering
spellingShingle DRNTU::Engineering
Toh, Jackson Yit Chuan.
Onset detection for drum loop classification
description The main objective of this report is to be able to perform drum beat detection through a semi-automatic way to produce the drum samples and then use these drum samples to produce a general classifier that is able to identify the drum beats in any music track. An algorithm was first introduced to perform onset detection on percussive sounds for a small period of music track of about six seconds. The user is involved in labeling which of these percussive sounds are the bass and snare drum, hence a semi-automatic way approach in drum beat detection. A small database of 21 music tracks was used in this project for performance evaluation. Performance evaluation of the onset detection algorithm is done by evaluating the average precision rate and recall rate of the algorithm for all the music tracks in the database. The percussive onset detection algorithm had an average precision rate of 0.714 and average recall rate of 0.953. After the onset detection, the next task was to train a classifier using the previous drum samples detected previously by the algorithm. There were two different scenarios in achieving this task. The first scenario was to classify the target vector into 2 classes with bass and snare drum, and the second scenario was to classify the target vector into 3 classes with the bass drum, snare drum and other onsets which are neither bass nor snare. The results of these two scenarios and possibilities for future work to solve this issue will be discussed in this report.
author2 Yap Kim Hui
author_facet Yap Kim Hui
Toh, Jackson Yit Chuan.
format Final Year Project
author Toh, Jackson Yit Chuan.
author_sort Toh, Jackson Yit Chuan.
title Onset detection for drum loop classification
title_short Onset detection for drum loop classification
title_full Onset detection for drum loop classification
title_fullStr Onset detection for drum loop classification
title_full_unstemmed Onset detection for drum loop classification
title_sort onset detection for drum loop classification
publishDate 2009
url http://hdl.handle.net/10356/17928
_version_ 1772828987415855104