Melody Recognition with Learned Edit Distances
Amaury Habrard, José Iñesta, David Rizo and Marc Sebban
In: S+SSPR 2008, december 2008, Orlando, USA.
In a music recognition task, the classiﬁcation of a new melody is often achieved by looking for the closest piece in a set of already known prototypes. The deﬁnition of a relevant similarity measure becomes then
a crucial point. So far, the edit distance approach with a-priori ﬁxed operation costs has been one of the most used to accomplish the task. In this paper, the application of a probabilistic learning model to both string and tree edit distances is proposed and is compared to a genetic algorithm cost ﬁtting approach. The results show that both learning models outperform ﬁxed-costs systems, and that the probabilistic approach is able to describe consistently the underlying melodic similarity model.