Department of Computer Science
Maynooth, Co. Kildare, Ireland
This paper presents a novel approach to the matching problem associated with timbre morphing. In particular a graph-theoretic technique, that of subgraph isomorphism, is applied to find correspondences between graph representations of a feature set for each timbre. The features are identified from Wigner distributions of the sounds using an adaptation of the McAulay-Quatieri sinusoidal analysis techniques. These features are then interpreted as nodes in graph representations of timbre. An audio morphing application can then be implemented by the application of spatial warping and linear interpolation of the Wigner distributions based on the correspondences established.
Download paper: [pdf] 353892 bytes