Beat-Based Rhythm Quantization of MIDI Performances

TL;DR

This paper introduces a transformer-based model for rhythm quantization of MIDI performances, leveraging beat and downbeat cues to produce metrically-aligned, human-readable scores, outperforming previous methods.

Contribution

It presents a novel beat-based preprocessing and a transformer architecture optimized for rhythm quantization, trained on piano and guitar data, achieving state-of-the-art results.

Findings

Outperforms existing models on MUSTER metric

Effective beat-based data representation

Applicable to piano and guitar MIDI performances

Abstract

We propose a transformer-based rhythm quantization model that incorporates beat and downbeat information to quantize MIDI performances into metrically-aligned, human-readable scores. We propose a beat-based preprocessing method that transfers score and performance data into a unified token representation. We optimize our model architecture and data representation and train on piano and guitar performances. Our model exceeds state-of-the-art performance based on the MUSTER metric.