On the Modeling of Time Information for Automatic Genre Recognition Systems in Audio Signals

The creation of huge databases coming from both restoration of existing analogue archives and new content is demanding fast and more and more reliable tools for content analysis and description, to be used for searches, content queries and interactive access. In that context, musical genres are crucial descriptors since they have been widely used for years to organize music catalogues, libraries and shops. Despite their use musical genres remain poorly defined concepts which make of the automatic classification problem a non-trivial task. Most automatic genre classification models rely on the same pattern recognition architecture: extracting features from chunks of audio signal and classifying features independently. In this paper, we focus instead on the low-level temporal relationships between chunks when classifying audio signals in terms of genre; in other words, we investigate means to model short-term time structures from context information in music segments to consolidate classification consistency by reducing ambiguities. A detailed comparative analysis of five different time modelling schemes is provided and classification results are reported for a database of 1400 songs evenly distributed over 7 genres.