Timbre Morphing Using the Wigner Time-frequency Distribution

Timbre morphing is a technique of music sound synthesis which combines existing sounds (timbres) to form a new sound. Musical timbre is defined as the temporal development of harmonics within a sound. We implement the Wigner time-frequency distribution as our representation of the temporal variation of harmonics. Similarity between input signals provides a basis for generating the Wigner distribution of the new timbre. In this paper we develop signal processing techniques for multi-signal timbre morphing using linear region-based interpolation and geometric warping. A cost function is used for the correspondence problem encountered in matching timbral features (peak: of attack etc.) in each sound. Recovery of the new timbre is achieved through additive synthesis. Due to interference terms, the Wigner distribution needs to be perturbed for the correct recovery of any signal. Results of synthesis are presented for synthetic signals. These provide a means to test and ensure a robust synthesis.