Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Accent Mandarin Dry-Vocal Singing Dataset: Benchmark for Singing Accent Recognition

Zihao Wang, Ruibin Yuan, Ziqi Geng, Hengjia Li, Xingwei Qu, Xinyi Li, Songye Chen, Haoying Fu, Roger B. Dannenberg, Kejun Zhang

TL;DR

This work tackles the lack of singing accent data by introducing MADVSD, a large Mandarin dry-vocal singing dataset with 670 hours from 4,206 speakers across nine regions, including both songs and phonetic vowel exercises. The authors benchmark singing accent recognition using a mix of audio-pretraining and speech-accent models, finding DIMNet to perform best overall and showing that dialect information can boost singing-accent tasks. They also analyze the relationship between dialect and accent and perform a vowel-level analysis to identify phonetic markers that distinguish regional singing accents. The dataset provides a valuable resource for benchmarking and phonetic studies in singing, with future directions including accent conversion and broader genre/geographic coverage.

Abstract

Singing accent research is underexplored compared to speech accent studies, primarily due to the scarcity of suitable datasets. Existing singing datasets often suffer from detail loss, frequently resulting from the vocal-instrumental separation process. Additionally, they often lack regional accent annotations. To address this, we introduce the Multi-Accent Mandarin Dry-Vocal Singing Dataset (MADVSD). MADVSD comprises over 670 hours of dry vocal recordings from 4,206 native Mandarin speakers across nine distinct Chinese regions. In addition to each participant recording audio of three popular songs in their native accent, they also recorded phonetic exercises covering all Mandarin vowels and a full octave range. We validated MADVSD through benchmark experiments in singing accent recognition, demonstrating its utility for evaluating state-of-the-art speech models in singing contexts. Furthermore, we explored dialectal influences on singing accent and analyzed the role of vowels in accentual variations, leveraging MADVSD's unique phonetic exercises.

Multi-Accent Mandarin Dry-Vocal Singing Dataset: Benchmark for Singing Accent Recognition

TL;DR

This work tackles the lack of singing accent data by introducing MADVSD, a large Mandarin dry-vocal singing dataset with 670 hours from 4,206 speakers across nine regions, including both songs and phonetic vowel exercises. The authors benchmark singing accent recognition using a mix of audio-pretraining and speech-accent models, finding DIMNet to perform best overall and showing that dialect information can boost singing-accent tasks. They also analyze the relationship between dialect and accent and perform a vowel-level analysis to identify phonetic markers that distinguish regional singing accents. The dataset provides a valuable resource for benchmarking and phonetic studies in singing, with future directions including accent conversion and broader genre/geographic coverage.

Abstract

Singing accent research is underexplored compared to speech accent studies, primarily due to the scarcity of suitable datasets. Existing singing datasets often suffer from detail loss, frequently resulting from the vocal-instrumental separation process. Additionally, they often lack regional accent annotations. To address this, we introduce the Multi-Accent Mandarin Dry-Vocal Singing Dataset (MADVSD). MADVSD comprises over 670 hours of dry vocal recordings from 4,206 native Mandarin speakers across nine distinct Chinese regions. In addition to each participant recording audio of three popular songs in their native accent, they also recorded phonetic exercises covering all Mandarin vowels and a full octave range. We validated MADVSD through benchmark experiments in singing accent recognition, demonstrating its utility for evaluating state-of-the-art speech models in singing contexts. Furthermore, we explored dialectal influences on singing accent and analyzed the role of vowels in accentual variations, leveraging MADVSD's unique phonetic exercises.

Paper Structure

This paper contains 27 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Dialects and Accents
  4. Dialect Datasets
  5. Regional Accent Datasets
  6. Dry Singing Vocal Datasets
  7. Recent Progress in AI Music Generation
  8. Dataset Construction of MADVSD
  9. Data Collection and Recording Setup
  10. Participant Recruitment and Training
  11. Dataset Composition and Content
  12. Formal Recording Procedures and Guidelines
  13. Accent Region Demarcation
  14. Recorded Data Post-Processing
  15. Data Availability and Privacy Protocol
  16. ...and 12 more sections

Figures (3)

  • Figure 1: The content of the Vowel Phonetic Exercises Protocol.
  • Figure 2: Performance of Accent Recognition Models
  • Figure 3: Vowel Analysis for YGSR and SSHR Accents.