Chord recognition with a Hidden Markov Model

Sašo Brus (2013) Chord recognition with a Hidden Markov Model. EngD thesis.

Preview

Abstract

In this paper a system for automatic chord estimation of an input song is presented. Our system is based on a Hidden Markov model – HMM. Visual representation of HMM elements is offered. Metric called Chromagram is used for evaluation of system states. Learn and evaluation processes are presented. Our system learns rules and performs evaluation on Isophonics musical database. Our system achieves 62% classification accuracy using 10-fold validation. Chord alphabet, used in our model, contains 25 chord states. We present reasons for achieved results and perform detailed estimation analysis. Our approach contains knowledge of music theory and psychoacoustics. All methods, used in our system are argued and compared with modern systems. Further, some options for improving classification accuracy are presented.

Item Type:

Thesis (EngD thesis)

Keywords:

chord, hmm, mirex, fft

Number of Pages:

Language of Content:

Slovenian

Mentor / Comentors:

Name and Surname	ID	Function
doc. dr. Matija Marolt	271	Mentor

Link to COBISS:

http://www.cobiss.si/scripts/cobiss?command=search&base=50070&select=(ID=10156884)

Institution:

University of Ljubljana

Department:

Faculty of Computer and Information Science

Item ID:

2197

Date Deposited:

21 Sep 2013 16:15

Last Modified:

02 Oct 2013 10:10

URI:

http://eprints.fri.uni-lj.si/id/eprint/2197

Actions (login required)

View Item