Computerized audio music recognition

Pao Yue-kong Library Electronic Theses Database

Computerized audio music recognition

 

Author: Ng, Sze-pan
Title: Computerized audio music recognition
Degree: M.Sc.
Year: 1994
Subject: Musical notation -- Data processing
Computer sound processing
Hong Kong Polytechnic -- Dissertations
Department: Dept. of Electronic Engineering
Pages: 1 v. (various pagings) : ill. ; 30 cm
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b1162847
URI: http://theses.lib.polyu.edu.hk/handle/200/3805
Abstract: Music offers a challenging array of recognition problems which arise in the context of entering musical data into a computer. As an art form, music is represented by the presence of many relationships that can treated mathematically, including pitch, loudness, tempo and timbre. There are also many nonmathematical elements such as tension, emotion or legato playing style which attribute to the difference between notated and played music. These elements combine to make music recognition a rich field of study. This report presents a monophonic, continuous piano music recognition system based on Time-Delay Neural Network (TDNN). The system is an alternative approach to traditional knowledge based systems which are designed to solve problems whose solutions have been made explicit. It employs TDNN to learn the characteristics of piano tones and performs continuous music recognition task based on the developed internal representations of the lower level acoustic cues. The system currently achieves 98 percents accuracy on the recognition of continuous tones from C4 to C5# (i.e. totally 14 tones), outperforming the traditional knowledge based recognition system using energy threshold or pitch detection segmentation method. TDNN is also found to be better in identifying the tone boundaries, especially when tones are embedded in background noise or overlapped owing to legato playing style. Another side product of the study is the developed algorithm to cater for the local minimum problem in error back propagation training. This algorithm solves the problem by dynamically adjusting the number of hidden units in the network and finally comes to an optimal configuration.

Files in this item

Files Size Format
b11628479.pdf 2.653Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information