Automatic classification of playing techniques in Guitar Pro songs
Bibliographic record
Abstract
Automatic guitar transcription has been an active research area for decades.Existing work in this area has mostly focused on estimating the onset, offset, and pitch of note events.Another important aspect of expressive guitar performance, the use of playing techniques, is less studied.The system presented in this thesis is designed to recognize five common playing techniques performed on the electric guitar: bend, vibrato, hammeron, pull-off, and slide.The system has three steps.For a given audio track, the monophony detector extracts all the monophonic segments, where most playing technique instances occur.Then, the note-event separator splits monophonic audio segments into note events.Finally, machine learning techniques are used to classify their playing techniques using various audio features such as pitch, spectral centroid, and mel-frequency cepstral coefficients.The presented system is trained and evaluated on a novel dataset of 379 synthesized guitar solo recordings, mostly in the genre of rock and metal, generated using publicly available Guitar Pro files collected from tablature websites.As an extended guitar tablature format, a Guitar Pro file encodes information about every note event in a guitar track, such as the timestamp, pitch, fingering position, and playing techniques.Using the Guitar Pro software, a realistic guitar audio track can be generated from the Guitar Pro file.Its corresponding annotation can also be obtained by programmatically decoding the Guitar Pro file.The end-to-end testing experiments showed that the presented system can effectively recognize the five playing techniques in synthesized guitar solos, with per-class f1 scores ranging from 71.7% to 89.2%.Major thanks are owed to my supervisor, Professor Ichiro Fujinaga, who introduced me to the world of music information retrieval (MIR) and generously shared knowledge and expertise during my time at McGill.Thank you for your warm welcome and helping me settle into Montral.Thank you for your advising on machine learning and MIR topics.Thank you for your generous funding and the opportunities to develop software projects for the lab.This thesis
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.002 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".