In musical instrument sound source separation, the temporal envelopes of the partials are correlated due to the physical constraints of the instruments. With this assumption, separation algorithms then exploit the similarities between the partial envelopes in order to group partials into sources. In this paper, we quantitatively investigate the partial temporal envelope similarities of a large database of instrument samples and develop weighting functions in order to model the similarities. These model partials then provide a reference to identify similar partials of the same source. The partial identification algorithm is evaluated in the separation of polyphonic mixtures and is shown to successfully discriminate between partials from different sources.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.