mlsanityData to 14 Oct 2026

BabyHuBERT: Multilingual Self-Supervised Learning for Segmenting Speakers in Child-Centered Long-Form Recordings

arXiv 2509.15001 ↗First posted 18 Sep 2025eess.AS cs.LG cs.SDcited by 5

Not on the map. Why?

Abstract (arXiv)

Child-centered daylong recordings are essential for studying early language development, but existing speech models trained on clean adult data perform poorly due to acoustic and linguistic differences. We introduce BabyHuBERT, a self-supervised speech model trained on 13,000 hours of multilingual child-centered recordings from 40+ languages. Evaluated on voice type classification, the task of identifying who produces speech and when in child-centered recordings (key child, other children, male, and female adults), BabyHuBERT-VTC achieves F1-scores from 55.0% to 76.1% across six corpora, consistently outperforming W2V2-LL4300 and HuBERT (pretrained on English daylongs and clean adult speech, respectively). Notable gains include 14.0 and 18.3 absolute F1 points over HuBERT on Vanuatu and Solomon Islands, demonstrating effectiveness on underrepresented languages. We share code and models to support researchers working with child-centered recordings across diverse linguistic contexts.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 15 of its 45 references that are in the map. First posted on arXiv 18 Sep 2025.

Builds on

15 of 45 references.

References · 15

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,905
  2. 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,342
  3. 2021WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processingcited by 1,657
  4. 2021SUPERB: Speech processing Universal PERformance Benchmarkcited by 666
  5. 2018Speaker Recognition from Raw Waveform with SincNetcited by 166
  6. 2021Systematic Inequalities in Language Technology Performance across the World's Languagescited by 120
  7. 2024mHuBERT-147: A Compact Multilingual HuBERT Modelcited by 66
  8. 2023TorchAudio 2.1: Advancing speech recognition, self-supervised learning, and audio processing components for PyTorchcited by 27
  9. 2020An open-source voice type classifier for child-centered daylong recordingscited by 14
  10. 2019Speaker detection in the wild: Lessons learned from JSALT 2019cited by 6
  11. 2023Towards Robust Family-Infant Audio Analysis Based on Unsupervised Pretraining of Wav2vec 2.0 on Large-Scale Unlabeled Family Audiocited by 6
  12. 2025Self-Supervised Models for Phoneme Recognition: Applications in Children's Speech for Reading Learningcited by 6
3 more
  1. 2025Fifteen Years of Child-Centered Long-Form Recordings: Promises, Resources, and Remaining Challenges to Validitycited by 4
  2. 2025Challenges in Automated Processing of Speech from Child Wearables: The Case of Voice Type Classifiercited by 3
  3. 2025Employing self-supervised learning models for cross-linguistic child speech maturity classificationcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026EgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Datacited by 0
  2. 2026Context-aware child-directed speech detection from long-form recordingscited by 0
  3. 2026Deriving Benchmarking Datasets from Long-Form Recordings: Challenges and Opportunitiescited by 0
  4. 2026Self-Supervised Speech Representations Track Spoken Language Convergence to Adult Models in Infants and Children Who Are Deaf/Hard-of-Hearingcited by 0
  5. 2026Cross-Linguistic Effects in Bilingual Phoneme BabyLMscited by 0