Drum Sound Recognition


İHoward Dodd


MSc Computing Science
Birkbeck College

University of London 1999
ISBN 0718716167

Contents

Abstract

1. Introduction

1.1. Project objective
1.2. Impact of samplers on modern music
1.3. Impact of computers in music production
1.4. Existing relevant software platforms
1.5. A new type of break beat processor
1.6. Preview of chapters


2. Background

2.1. Background of problem
2.2. Solving the problem

3. Description of methods

3.1. Input and Output
3.2. Pre-processing
3.3. Neural Networks
3.4. Multi-Layer Perceptrons (MLP)
3.5. Kohonen Self Organising Maps (SOM)
3.6. Post classification

4. Requirements

4.1. Overview of Requirements
4.2. User Interface
4.3. Pre-processing
4.4. Neural Networks
4.5. Post classification

5. Design & Implementation

5.1. Overview of software structure
5.2. Pre-processing
5.3. Neural networks
5.4. Post classification

6. Analysis of Results

6.1. Training of the Feed forward MLPs
6.2. Recurrent MLPs
6.3. Kohonen SOM
6.4. Post classification

7. Discussion and Conclusions

7.1. Overall performance of project
7.2. Spectral analysis
7.3. Noise reduction
7.4. The Neural Networks

8. Further Developments
8.1. Recognition developments
8.2. Improvements on the existing structure
8.3. An overview of the finished application


References


Appendices

A Classes, structures and Global Functions
Format of data within .rcy files