Toward Part-Aware Choral Transcription with singing voice assignment

2026-10-07

AI Summary

This study addresses the challenge of voice overlap and the difficulty of separating independent SATB tracks in choral music transcription by proposing the first end-to-end neural framework that jointly models note transcription and voice assignment. Methodologically, it introduces a novel voice-aware transcription paradigm utilizing neural networks to predict note onset and offset frames. Furthermore, it incorporates vocal range priors and ordered continuity to construct a structured training objective, combined with a voice presence estimation strategy to achieve precise transcription. Experimental results demonstrate that the proposed method improves the voice-aware F1 score on the YouChorale dataset by 36.4% over baseline models, significantly enhancing the separation performance of individual tracks from polyphonic choral audio.

Abstract

Single-instrument automatic music transcription (AMT) has advanced substantially, yet choral applications require soprano, alto, tenor, and bass (SATB) to be transcribed as separate parts. Recent note-level choral AMT instead produces a single merged note track, limiting rehearsal, education, and score reconstruction. To address this limitation, we introduce Part-aware Choral Transcription (PawCT), to our knowledge the first end-to-end neural framework that identifies active SATB parts from choral audio and transcribes each into a separate note-level track. PawCT combines part-specific onset, offset, and frame prediction with part-presence estimation, union-level supervision, and structured training targets using a range prior (RP) based on SATB pitch ranges and its ordered-continuity (OC) extension, which adds within-part melodic continuity and cross-part pitch ordering. On YouChorale, PawCT-RP-OC achieves a macro part-aware note F1 of 0.225 at a 50-ms onset tolerance, outperforming an adapted choral baseline (0.165) by 36.4% relative and a two-stage post-hoc assignment pipeline (0.175). Its part-agnostic variant, PagCT, achieves a 50-ms onset F1 of 0.382, compared with 0.237 for the previous state-of-the-art choral AMT model. Cross-dataset evaluations on CSD and Cantoria further assess performance under dataset shift. These results demonstrate the benefit of jointly modeling note transcription and vocal-part assignment. Code and demos are available at https://hanyu-meng.github.io/Paw_Choral_AMT_Demo/.

Problem

• Choral Transcription
• Automatic Music Transcription
• Singing Voice Assignment
• Part-aware
• SATB

Innovation

• Choral Transcription
• End-to-end Framework
• Singing Voice Assignment
• Range Prior
• Ordered-Continuity