Performance Analysis of a Convolutional Transformer-Based Automatic Chord Recognition Model for Extended Chord Qualities

• Daniel Antonio de Jesus Melo UFPB
• Pedro Augusto Gomes Medeiros UFPB
• Luigi Emanuel Martins Schmitt UFPB
• Telmo M. Silva Filho University of Bristol
• Yuri de Almeida Malheiros Barbosa UFPB
• Thaís Gaudencio do Rêgo UFPB

DOI: https://doi.org/10.5753/sbcm.2025.13240

Resumo

Automatic Chord Recognition (ACR) has seen significant advancements with the advent of deep learning, particularly with convolutional transformer architectures. However, a detailed performance analysis, especially concerning their ability to recognize chord qualities, remains an area for investigation. This paper presents an analysis of a convolutional transformer-based automatic chord recognition model for extended chord qualities. We evaluate its accuracy on three benchmark datasets—McGill-Billboard, RWC-Pop, and JAAH—, focusing on root notes and specific chord qualities (maj, min, maj7, min7, 7, hdim7, and 7(b9)). Using annotations from the original ground truth files and the CHOCO (Chord Corpus) as a reference, with a duration-weighted evaluation methodology, we conduct an assessment of the model’s capabilities. Our results reveal that while the model achieves high accuracy on basic triads, its ability to correctly identify extended chords drops in comparison. We conclude that the model’s limitations stem from the underrepresentation of extended chords in common training corpora, underscoring the critical need for more balanced and harmonically diverse datasets to potentially mitigate the reported issues.

Referências

Matt McVicar, Raul Santos-Rodriguez, Yizhao Ni, and Tijl De Bie. Automatic chord estimation from audio: a review of the state of the art. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(2):556–575, 2014.

Filip Korzeniowski, David R. W. Sears, and Gerhard Widmer. A large-scale study of language models for chord prediction. arXiv preprint arXiv:1804.01849, 2018.

Jonggwon Park, Kyoyun Choi, Sungwook Jeon, Dokyun Kim, and Jonghun Park. A bi-directional transformer for musical chord recognition. arXiv preprint arXiv:1907.02698, 2019.

Muhammad Waseem Akram, Stefano Dettori, Valentina Colla, and Giorgio Carlo Buttazzo. Chordformer: a conformer-based architecture for large-vocabulary audio chord recognition. arXiv preprint arXiv:2502.11840, 2025.

Masataka Goto, Hiroki Hashiguchi, Takuichi Nishimura, and Ryuichi Oka. Rwc music database: popular, classical, and jazz music databases. In Proceedings of the 3rd International Conference on Music Information Retrieval, 2002.

John Burgoyne, Jonathan Wild, and Ichiro Fujinaga. An expert ground truth set for audio chord recognition and music analysis. In Proceedings of the 12th International Society for Music Information Retrieval Conference, pages 633–638, 2011.

Eric Humphrey, Justin Salamon, Oriol Nieto, Jon Forsyth, Rachel Bittner, and Juan Bello. Jams: a json annotated music specification for reproducible mir research. In Proceedings of the 15th International Society for Music Information Retrieval Conference, 2014.

Jacopo De Berardinis, Albert Meroño-Peñuela, Andrea Poltronieri, and Valentina Presutti. Choco: a chord corpus and a data transformation workflow for musical harmony knowledge graphs. Scientific Data, 10(1):641, 2023.

Vsevolod Eremenko, Emir Demirel, Baris Bozkurt, and Xavier Serra. Audio-aligned jazz harmony dataset for automatic chord transcription and corpus-based research. Zenodo, 2018.

Johan Pauwels and Geoffroy Peeters. Evaluating automatically estimated chord sequences. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, pages 749–753, 2013.

Luke Rowe and George Tzanetakis. Curriculum learning for imbalanced classification in large vocabulary automatic chord recognition. In Proceedings of the 22nd International Society for Music Information Retrieval Conference, 2021.

Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th Annual International Conference on Machine Learning, pages 41–48. ACM, 2009.

Christopher Harte, Mark Sandler, Samer Abdallah, and Emilia Gómez. Symbolic representation of musical chords: a proposed syntax for text annotations. In Proceedings of the 6th International Conference on Music Information Retrieval, pages 66–71, 2005.