Audio to Score Alignment for Educational Software Interim Presentation for the MSc Personal Project Antoine Gomas supervised by Dr. Tim Collins 22nd of June, 2007
Agenda Introduction Objectives Review & Innovation Work Conclusion Achievements so far Planned work Conclusion
Audio to score alignment? Associate Notes in a score Timing points in a recording Example
Project objectives Implement a monophonic audio to score alignment algorithm Evaluate characteristics of the performance Design a learning interface to help music students improve their performance
Review (1) Previous work Algorithms already exist Similar to Spoken Language Processing Application: musicology Professional recordings
Review (2) Previous work (continued) Dynamic Time Warping Few parameters Heavy Low flexibility Hidden Markov Models Very flexible Large number of parameters (training)
Review (3) Innovation Apply to educational software Requires modifications & new functionalities Cope with errors Detect errors
Work First system Results so far Work plan for the next two months
First system (1)
First system (2) First “working” version Attack, Sustain, Silence Uses Dynamic Time Warping
First Results Works for simple cases: Good at rhythm recuperation Short performances Clean synthetic music Good at rhythm recuperation Requires correct pitches
Planned work Switch to HMMs Design learning interface Lower computing requirements More flexible to recover from student’s errors Design learning interface Thorough review about design standards No implementation expected
Conclusion Promising first results HMMs risky but interesting Challenging project
Thank you for listening Any questions ?