Three-Dimensional Mapping of High-Level Music Features for Music Browsing

Abstract

The increased availability of musical content comes with the need of novel paradigms for recommendation, browsing and retrieval from large music libraries. Most music players and streaming services propose a paradigm based on content listing of meta-data information, which provides little insight on the music content. In services with huge catalogs of songs, a more informative paradigm is needed. In this work we propose a framework for music browsing based on the navigation into a three-dimensional (3-D) space, where musical items are placed as a 3-D mapping of their high-level semantic descriptors. We conducted a survey to guide the design of the framework and the implementation choices. We rely on state-of-the-art techniques from Music Information Retrieval to automatically extract the high-level descriptors from a low-level representation of the musical signal. The framework is validated by means of a subjective evaluation from 33 users, who give positive feedbacks and highlight promising future developments especially in virtual reality field.