Neural model transcribes choral music into separate vocal parts
Toward Part-Aware Choral Transcription with singing voice assignment
Sound
Summary
Transcribing choral music into individual singing parts like soprano, alto, tenor, and bass is hard because most automatic systems merge all notes into one track. The authors created a new neural system called PawCT that not only detects notes but also assigns them to the correct vocal parts from choral recordings. Their model uses special techniques to understand pitch ranges and order of parts, improving accuracy compared to previous methods. This can help in music practice, teaching, and making digital sheet music from choral performances.
What this means in practice
- •For choral conductors: Automatically produce separate vocal part transcriptions to aid rehearsal and score analysis of choir performances.
- •For music software developers: Integrate end-to-end choral transcription tools that output part-assigned note data for digital score reconstruction and education apps.
Authors
Hanyu Meng, Zhanhong He, Zixun Guo, Yaolong Ju
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/.