Supplementary materials for ISMIR 2022 LBD submission: https://archives.ismir.net/ismir2022/latebreaking/000032.pdf
- Chords played: C-major, G-major, F-major, A-minor
- Duration: 10 examples with a total duration of 5 minutes
- Motion signal synchronized to audio signals
- Labels for up and down strumming events
The files can be found in the "dataset" subdirectory.
The motion signal is recording using affordable hardware setup. The IMU used is an GY-521 6-axis gyros-cope and accelerometer which is mounted to the back of the hand of the guitar player.
- xyz_line.wav -> Line Audio Recording using Behringer GUITAR 2 USB cable
- xyz_phone.wav -> Smartphone Audio Recording
- xyz.csv -> Motion signal recording stored in csv format with time and value columns
- xyz.strums -> Strumming annotations
A baseline method has been implemented which uses audio for strumming event detection and the hand motion signal for strumming direction classification.
python scripts/evaluate.py
| Strumming Class | Precision | Recall | F1 Score |
|---|---|---|---|
| Down strums | 89.88% | 94.53% | 91.94% |
| Up strums | 79.14% | 93.07% | 84.65% |
If you use this code or dataset in your research, please cite us via the following BibTeX:
@article{murgul2022multimodal,
title={A Multimodal Approach to Acoustic Guitar Strumming Action Transcription},
author={Murgul, Sebastian and Heizmann, Michael},
journal={Proc. ISMIR Late-Breaking and Demo},
year={2022}
}