University of Surrey

Test tubes in the lab Research in the ATI Dance Research

Timbre remapping through a regression-tree technique

Stowell, D and Plumbley, MD (2010) Timbre remapping through a regression-tree technique

Full text not available from this repository.


We consider the task of inferring associations between two differently-distributed and unlabelled sets of timbre data. This arises in applications such as concatenative synthesis/ audio mosaicing in which one audio recording is used to control sound synthesis through concatenating fragments of an unrelated source recording. Timbre is a multidimensional attribute with interactions between dimensions, so it is non-trivial to design a search process which makes best use of the timbral variety available in the source recording. We must be able to map from control signals whose timbre features have different distributions from the source material, yet labelling large collections of timbral sounds is often impractical, so we seek an unsupervised technique which can infer relationships between distributions. We present a regression tree technique which learns associations between two unlabelled multidimensional distributions, and apply the technique to a simple timbral concatenative synthesis system. We demonstrate numerically that the mapping makes better use of the source material than a nearest-neighbour search. © 2010 Dan Stowell et al.

Item Type: Conference or Workshop Item (UNSPECIFIED)
Divisions : Surrey research (other units)
Authors :
Stowell, D
Date : 1 January 2010
Depositing User : Symplectic Elements
Date Deposited : 17 May 2017 13:21
Last Modified : 23 Jan 2020 18:29

Actions (login required)

View Item View Item


Downloads per month over past year

Information about this web site

© The University of Surrey, Guildford, Surrey, GU2 7XH, United Kingdom.
+44 (0)1483 300800