Overview

Analysis of piano pedalling techniques from sensor, audio and midi data

============= WARNING================ =====================================
This project is now mainly used for backup,
i.e., not user-friendly downloadable/reviewable.

Please check the my github repo for updates: ======================================

analysis-sensor-data

bela4dataCapture: The measurement system can capture the pedalling gesture by near-field optical reflectance sensing and simultaneously record sensor data and piano sound through an embedded platform (Bela) under normal playing conditions.

py4sensordata: Using the sensor data collected from the system, the recognition is comprised of two separate tasks, onset/offset detection and classification.

matlab4GUI: Recognition results can be represented by customised pedalling notations and visualised in an audio based score following application.

analysis-audio-data

midi4disklavier & sliceAudio: Since there is no publicly available dataset for piano pedalling technique detection across different tones and velocities, we build our own dataset with isolated notes played without the sustain pedal as well as using different pedalling techniques. Specifications for notes with different pedalling conditions were encoded as standard MIDI files and then audio was generated using a reproducing piano. This provided fully-automatic and reliable annotation for our audio dataset.

py4audio: python codes for pedalling techniques detection from audio recordings (work in progress)

analysis-midi-data

Notations of piano pedalling technique in the music score are usually lacking in detail: they provide boundary locations of pedalling techniques, but do not indicate what musical attribute prompts the pedalling change. Understanding this relationship would be useful for musicology and piano pedagogy. We propose to model how musically-motivated features correlate with pedalling transitions. Our aim is to employ this model as prior information for the detection of pedal onsets and offsets from audio recordings. (work in progress)

Related publications

B. Liang, G. Fazekas, A. P. McPherson, and M. B. Sandler, “Piano Pedaller: A Measurement System for Classification and Visualisation of Piano Pedalling Techniques,” in Proceedings of the International Conference on New Interfaces for Musical Expression (NIME), 2017.
[More Details] [BIBTEX] [URL (ext.)]
B. Liang, G. Fazekas, and M. B. Sandler, “Recognition of Piano Pedalling Techniques Using Gesture Data,” in Proceedings of the 12th International Audio Mostly Conference on Augmented and Participatory Sound and Music Experiences, 2017.
[More Details] [BIBTEX] [URL (ext.)]
B. Liang, G. Fazekas, and M. B. Sandler, “Detection of Piano Pedaling Techniques on the Sustain Pedal,” in Audio Engineering Society Convention 143, 2017.
[More Details] [BIBTEX] [URL (ext.)]

Members

Manager: Beici Liang, Gyorgy Fazekas