A Music Recommendation System based on Semantic Audio Segments Similarity

A. Bozzon, G. Prandi, G. Valenzise, and M. Tagliasacchi (Italy)


Content-based Multimedia Retrieval, Music Recommenda tion, Genre Classification


In this paper we propose a novel approach for content based music recommendation. The main innovation of the proposed technique consists of a similarity function that, instead of considering entire songs or their thumbnail rep resentations, analyzes audio similarities between semantic segments from different audio tracks. The rationale of our idea is that a song similarity and recommendation tech nique, to be more meaningful to the user from a seman tic point of view, may evaluate and exploit similarities on semantic units between audio tracks. Our similarity algo rithm consists of two main stages: the first step performs segmentation of the song in semantic parts. The latter as signs a similarity and recommendation score to a pair of songs, by computing the distance between the represen tations of their segments. To assign the global similar ity and recommendation score, we consider a consistent subset of all the inter-segment distances. By adopting a graph-bases framework, we propose a graph-reduction al gorithm on weighted edges that connect segments of dif ferent songs to optimize the similarity score with respect to our recommendation goal. Experiments conducted on a database of 200 audio tracks of various authors and genres show promising results.

Important Links:

Go Back