Huw Cheston

Huw Cheston

he/him

Music computing & data science

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

portfolio

Identifying hip-hop samples with deep learning

Sampling involves reusing recorded music or sounds from another source and is common in genres like hip-hop and rap. We trained a deep learning model to identify samples in a large music catalog from Spotify and developed an interactive web application to explore the predictions it made.

Jazz trio database

A dataset of 45 hours of commercial jazz recordings annotated using an automated signal processing pipeline

publications

Rhythmic Qualities of Jazz Improvisation Predict Performer Identity and Style in Source-Separated Audio Recordings.

Published in Royal Society Open Science, 2024

We demonstrate that a supervised learning model trained solely on rhythmic features extracted from 300 source-separated audio recordings of jazz pianists was capable of identifying the performer in 59% of cases, over five times better than chance.

Recommended citation: Cheston, H., Schlichting, J. L., Cross, I., & Harrison, P. M. C. (2024). Rhythmic Qualities of Jazz Improvisation Predict Performer Identity and Style in Source-Separated Audio Recordings. Royal Society Open Science, 11(11). https://doi.org/10.1098/rsos.240920 https://doi.org/10.1098/rsos.240920

Jazz Trio Database: Automated Timing Annotation of Jazz Piano Trio Recordings Processed Using Audio Source Separation

Published in Transactions of the International Society for Music Information Retrieval, 2024

We introduce the Jazz Trio Database, a dataset of 45 hours of jazz piano trio recordings with automatically generated annotations for every performer (piano soloist, bass and drums accompaniment) in the ensemble.

Recommended citation: Cheston H, Schlichting JL, Cross I, & Harrison PMC. Jazz Trio Database: Automated Annotation of Jazz Piano Trio Recordings Processed Using Audio Source Separation. Transactions of the International Society for Music Information Retrieval. 2024; 7(1), 144–158. https://doi.org/10.5334/tismir.186

Trade-offs in Coordination Strategies for Duet Jazz Performances Subject to Network Delay and Jitter

Published in Music Perception, 2024

We show that five duos of professional jazz musicians adopt diverse strategies when confronted by the difficulties of coordinating performances over a network — difficulties that are not exclusive to networked performance, but are also present in other situations (such as when coordinating performances over large physical spaces).

Recommended citation: Cheston H, Cross I, & Harrison PMC. Trade-offs in Coordination Strategies for Duet Jazz Performances Subject to Network Delay and Jitter. Music Perception. 2024; 42(1), 48–72. https://doi.org/10.1525/mp.2024.42.1.48

Automatic Identification of Samples in Hip-Hop Music via Multi-Loss Training and an Artificial Dataset

Preprint published on arXiv, 2025

We show that a convolutional neural network trained on an artificial dataset can identify real-world samples in commercial hip-hop music. Our model achieves 13% greater precision on real-world instances of sampling than a fingerprinting system using acoustic landmarks (Shazam-style), and can recognize samples that have been both pitch shifted and time stretched.

Recommended citation: Cheston H, Van Balen J, & Durand S. Automatic Identification of Samples in Hip-Hop Music via Multi-Loss Training and an Artificial Dataset. arXiv. 2025; arXiv:2502.06364 [cs.SD]. https://doi.org/10.48550/arXiv.2502.06364

Understanding Jazz Improvisation Style with Explainable Music Performer Identification Models

Preprint published on arXiv, 2025

We construct a series of models to identify jazz performers from audio recordings, culminating in a multi-input convolutional neural network architecture that achieves state-of-the-art prediction accuracy (91\% success in identifying twenty famous jazz performers) with an interpretable structure that allows its predictions to be explained in terms of four fundamental musical domains (melody, harmony, rhythm, and dynamics).

Recommended citation: Cheston H, Bance R, & Harrison PMC. Understanding Jazz Improvisation Style with Explainable Music Performer Identification Models. arXiv. 2025; arXiv:COMING_SOON [cs.SD]. COMING_SOON

talks

teaching

Supervisor and Guest Lecturer, Topics in Music & Science

Undergraduate course, University of Cambridge, Faculty of Music, 2022

  • Designed & delivered small-group supervisions for Undergraduate students on: (i) analysing audio recordings, (ii) introduction to programming in Python and R, (iii) visualising and simulating data, (iv) overall course revision
  • Designed & delivered lecture on analysing audio recordings, involving introduction and demonstration of basic Music Information Retrieval concepts.

Supervisor and Guest Lecturer, Topics in Music & Science

Undergraduate course, University of Cambridge, Faculty of Music, 2023

  • Designed & delivered small-group supervisions for Undergraduate students on: (i) analysing audio recordings, (ii) designing a research proposal, (iii) overall course revision
  • Designed & delivered lecture on analysing audio recordings, involving introduction and demonstration of basic Music Information Retrieval concepts.