Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.534917
Title: An iterative, residual-based approach to unsupervised musical source separation in single-channel mixtures
Author: Siamantas, Georgios
Awarding Body: University of York
Current Institution: University of York
Date of Award: 2009
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
This thesis concentrates on a major problem within audio signal processing, the separation of source signals from musical mixtures when only a single mixture channel is available. Source separation is the process by which signals that correspond to distinct sources are identified in a signal mixture and extracted from it. Producing multiple entities from a single one is an extremely underdetermined task, so additional prior information can assist in setting appropriate constraints on the solution set. The approach proposed uses prior information such that: (1) it can potentially be applied successfully to a large variety of musical mixtures, and (2) it requires minimal user intervention and no prior learning/training procedures (i.e., it is an unsupervised process). This system can be useful for applications such as remixing, creative effects, restoration and for archiving musical material for internet delivery, amongst others. Here, specific priors include that the signal contains detectable musical events, with characteristic partial structures, often assumed to be harmonic. The harmonicity cue is incorporated by employing an adapted and extended frame-based multiF0 estimator for identifying the sources. This acts as a front-end to a source estimation and extraction stage. Further, an iterative procedure is introduced between the two stages, enabling improved performance via increased adaptivity to signal content: this novel approach becomes possible by exploiting a residual signal. Experimental results show that the proposed residual-based method achieves better average performance compared to alternative methods in terms of source separation and multiF0 estimation on a range of mixtures of varying complexity. Unmodelled content of the separated mixture will appear in the residual, which can be exploited further. In particular, a novel onset detection technique is proposed that works entirely with the residual. Considering its simplicity, the technique shows promising results compared to two existing methods that do not use the residual.
Supervisor: Szymanski, John E. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.534917  DOI: Not available
Share: