Code repository for ICASSP 2015 & ISMIR 2015 submission on repeated themes discovery from audio recordings and MIDI with VMO. Created and maintained by Cheng-i Wang. Implementation by Cheng-i Wang and Jennifer Hsu.

email:chw160@ucsd.edu

• vmo
• Numpy, Scipy, matplotlib
• mir_eval

• bin

  • JKUPDD_experiment.py

  The experiment script generating pattern discovery algorithm output files.

  • JKUPDD_time.py

  The timing script.

• notebook/JKUPDD_demo.ipynb

  A python notebook demonstrating how vmo is used to find repeated themes from audio recordings.

• JKUPDD-Aug2013

  The MIREX dataset.

• estimation_results

  The experiment results using the scripts.

• supplementPlots

  Folders corresponding to the five music pieces from the dataset have supplement plots not shown in the submitted paper.

  • irProfile The information rate of different VMO created by different threshold values. The vmo with the highest information rate is used to find the repeated themes.

  • jkupdd_patterns Beat-synchronous chromagram, ground truth and the algorithm outputs.

  • chromagramAndRecurrencePlot Beat-synchrounous chromagram and a recurrence plot extracted directly from the VMO symbolization results. The recurrence plot is depicted only for visualization purpose, and is not used during the pattern finding algorithm. The pattern finding algorithm relies only on the suffix structure and values of longest repeated suffices from VMO.