Using the SDIF Sound Description Interchange Format for Audio Features

We present a set of extensions to the Sound Description Interchange Format (SDIF) for the purpose of storage and/or transmission of general audio descriptors. The aim is to allow portability and interoperability between the feature extraction module of an audio information retrieval application and the remaining modules, such as training, classification or clustering. A set of techniques addressing the needs of short-time features and temporal modeling over longer windows are proposed, together with the mechanisms that allow further extensions or adaptations by the user. The paper is completed by an overview of the general aspects of SDIF and its practical use by means of a set of existing programming interfaces for, among others, C, C++ and Matlab.