Pattern Mining in Discrete Time Series and Application to Music Mining

Authors:
Olivier Lartillot
Affiliations:
Music Department, University of Jyväskylä, Finland
Venue:
Proceedings of the 2006 conference on Advances in Intelligent IT: Active Media Technology 2006
Year:
2006

Citing 7
Cited 0

Formal Concept Analysis: Mathematical Foundations

Formal Concept Analysis: Mathematical Foundations
Mining Sequential Patterns

ICDE '95 Proceedings of the Eleventh International Conference on Data Engineering
Mining Partially Periodic Event Patterns with Unknown Periods

Proceedings of the 17th International Conference on Data Engineering
InfoMiner+: Mining Partial Periodic Patterns with Gap Penalties

ICDM '02 Proceedings of the 2002 IEEE International Conference on Data Mining
Efficient Mining of Partial Periodic Patterns in Time Series Database

ICDE '99 Proceedings of the 15th International Conference on Data Engineering
Efficient Algorithms for Mining Closed Itemsets and Their Lattice Structure

IEEE Transactions on Knowledge and Data Engineering
Discovery of Time-Series Motif from Multi-Dimensional Data Based on MDL Principle

Machine Learning

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new methodology for automated extraction of repeated patterns in discrete time series data is presented, dedicated in particular to the discovery of repeated motives in symbolic music representations, such as MIDI files. The basic principle of the approach consists in a search for closed patterns in a multi-dimensional parametric space. The pattern description is further reduced through a lossless pruning of the sequence description. A notion of cyclic pattern is introduced, enabling an adapted filtering of another form of combinatorial redundancy caused by successive repetitions of patterns. The use of cyclic patterns implies a necessary chronological scanning of the musical sequence. This study shows therefore that automated analysis of music cannot rely on simple mathematical or statistical approaches, but needs rather a complex and detailed modeling of the cognitive system ruling listening processes. The resulting algorithm is able to offer for the first time compact and relevant motivic analyses of simple monodies, and may therefore be applied to automated indexing of symbolic music databases. Numerous additional mechanisms need to be added in order to consider all aspects of music expression, including polyphony and complex musical transformations.