Skip to content

Expressive guitar technique classifier. Code to train and test multiple classifiers (and perform grid search)

License

Notifications You must be signed in to change notification settings

domenicostefani/ExpressiveGuitar-TechniqueClassifier

Repository files navigation

Expressive Guitar Technique classifier

Ph.D. research project of Domenico Stefani
The Jupyter notebook loads a dataset of feature vectors extracted from pitched and percussive sounds recorded with many acoustic guitars.


The techniques/classes are:

  1. Kick (Palm on lower body)
  2. Snare 1 (All fingers on lower side)
  3. Tom (Thumb on higher body)
  4. Snare 2 (Fingers on the muted strings, over the end of the fingerboard)

  1. Natural Harmonics (Stop strings from playing the dominant frequency, letting harmonics ring)
  2. Palm Mute (Muting partially the strings with the palm of the pick hand)
  3. Pick Near Bridge (Playing toward the bridge/saddle)
  4. Pick Over the Soundhole (Playing over the sound hole) (NEUTRAL NON-)TECHNIQUE

Content of the repository

  • data/: Folder containing the links to the feature dataset files. Download them in the folder with the links file.
  • phase3results/: Results of Experiment1 for a scientific paper submitted to IEEE/ACM Transactions on Audio, Speech, and Language Processing (TASLP).
  • convert_to_script.py : Script to convert the Colab/Jupyter notebook to a Python script.
  • expressive-technique-classifier-phase3.ipynb : Jupyter notebook with the code to train and test the classifier.
  • guitarists_touch.ipynb: Jupyter notebook with the code to train and test the classifier for Experiment 3.
  • run_grid_search.py : Script to run a grid search on the classifier.

Contact Domenico Stefani for any issues with running the code to repeat the experiments.

domenico[dot]stefani[at]unitn[dot]it
work.domenicostefani.com

About

Expressive guitar technique classifier. Code to train and test multiple classifiers (and perform grid search)

Topics

Resources

License

Stars

Watchers

Forks