Structuring and Querying Documents in an Audio Database Management System

The growing digitization of multimedia content must be supported by a set of tools to manipulate them, and especially to query them. This is one of the major goals of an audio DBMS. Yet, existing work related to audio documents, e.g., radio or television archives, often leave the DBMS question open. In this paper, we lay the foundations for integrating audio into a general purpose DBMS, in the form of an audio abstract data type, along with its properties and associated operators. This contribution is coupled with an unsupervised statistically-founded speaker-based partitioning technique. For each of these two aspects, the paper underlines the practical interest and some technical difficulties. Also, some query examples introduce the problem of the complexity of the querying expressions as well as of time complexity.