Iterative Separation of Note Events from Single-Channel Polyphonic Recordings

Author: Alejandro Delgado Castro

A thesis submitted at the University of York for the degree of Doctor of Philosophy

The aim of this webpage is to present audio examples associated with specific sections or experiments within the thesis, thus provinding a personal listening experinece to the reader, or to any other person interested in this matter. In every case, the original unmixed sources are presented along with the input mixture and the estimated sources that were obtained with the methods investigated. Notice that an uncompressed audio format (wav) has been used in every example, in order to provide the highest audio quality and to allow further comparison of the results.

Chapter 4 - An Iterative Note Event-based Multipitch Estimator

Audio experpts associated with the four sets of pitch trajectories presented in Figure 4.11 (V = Violin, C = Clarinet, S = Saxophone, and B = Bassoon).

Chapter 5 - Semi-supervised Source Separation based on Clustering of Note Events

Individual piano notes presented in Figure 5.5.

Individual violin notes presented in Figure 5.6.

Time domain subtraction applied to a set of audio mixtures in Figure 5.16 (V = Viola, C = Clarinet, S = Saxophone, and B = Bassoon).

Comparison between the original and separated notes in Figure 5.17.

Estimated note events in a mixture of viola an clarinet shown in Figure 5.19.

Comparison between the original and separated sources in Figure 5.20.

Separation of several synchronous violin notes using two different extraction methods (Table 5.2).