Voice Separation: A 15-minute Introduction My name is Denis Lebel and I will talk about Interactive Rendering of Suggestive Contours with Temporal Coherence. presented by Denis Lebel

Presentation Outline Introduction Voice Separation Techniques Voice Separation Systems Conclusion References To start off this presentation, we will look at an example of suggestive contours vs true contours and clarify the terminology I will be using throughout my presentation. I will then give a brief overview of suggestive contours, so you get a better idea of what they actually are. Then, we’ll move the motivation behind the work of this paper and cover the various contributions by the author. I will end this presentation by giving you future challenges for suggestive contours. If time permits it, you will also have a chance to watch a live demonstration of suggestive contours… One more thing: feel free to ask questions if you don’t understand and I’ll do my best to answer or will refer you to a more adequate source of information. MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Introduction Purpose Idea 2 different contexts Transcription of low-level musical data into score notation Theme finding and music analysis Idea Separation of notes into voices with possible chords (in a polyphonic context) 2 different contexts Explicit Polyphony Multiple notes sounding at one instant Assuming no single voice can produce 2 notes simultaneously Implicit Polyphony At most one note sounding at any instant MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Techniques Split Point Separation Idea: Split pitch range into disjoint intervals Problem: Works only on non-overlapping voices Notes: One of the simplest methods Used in most commercial systems MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Techniques Rule-Based Approach Idea: Take advantage of the voice-leading rules used by composer Rules examples: Polyphonic motion Succeeding notes intervals Problems: Many such rules, specific to composer Errors occur with major overlapping of voices Note: Better approach than split point MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Techniques Local Optimization Approach Idea: Using a heuristic algorithm Iterative process that finds the best solution from a given set at each step Problem: Not meant to find the correct voice separation but rather to provide reasonable solutions in different contexts Note: Complex approach but seems to give better results MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Techniques Others Contig Mapping Approach Same-Voice Predicate (learned decision tree) along with voice-numbering algorithm … MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Systems VoiSe – University of Massachusetts Amherst (Kirlin 2005) VoSA – University of Southern California (Chew 2004) Melisma Music Analyzer – Carnegie Mellon University (Sleator and Temperley 2001) MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Systems VoSA Interface Screenshot – from VoSA website MUMT-611: Music Information Acquisition, Preservation, and Retrieval

Conclusion Despite the different existing techniques, there is still no perfect solution to voice separation. MUMT-611: Music Information Acquisition, Preservation, and Retrieval

References Kirlin, P., and P. Utgoff. 2005. VoiSe: Learning to segregate voices in explicit and implicit polyphony. Proceedings of the International Conference on Music Information Retrieval. 552–7. Chew, E., and X. Wu. 2004. Seperating voices in polyphonic music: A contig mapping approach. Proceedings of the International Symposium on Computer Music Modeling and Retrieval. 1–20. Kilian, J., and H. Hoos. 2002. Voice separation: A local optimisation approach. Proceedings of the International Conference on Music Information Retrieval. 39–46. MUMT-611: Music Information Acquisition, Preservation, and Retrieval