VibraVox: French Speech Corpus via Body-Conduction Transducers

VibraVox contains between 10,000 and 100,000 French speech recordings captured using body-conduction transducers. Developed by Cnam-LMSSC and documented in Arxiv 2407.11828, this dataset provides a specialized audio-text corpus for speech processing research. It includes expert-generated and crowdsourced annotations for various audio-centric machine learning tasks.

Use Cases

• Automatic Speech Recognition (ASR) using body-conducted audio signals and text transcripts
• Speaker Identification based on the unique acoustic signatures of body-conduction transducers
• Audio-to-Audio synthesis to map or enhance body-conducted speech into standard audio formats

Strengths

• Scale of 10,000 to 100,000 records
• Expert-generated annotations for high-quality labeling
• CC BY 4.0 permissive license for research and commercial use

Limitations

• Restricted to the French language
• Acoustic properties are specific to body-conduction hardware and may not generalize to standard air-conduction microphone models

Provenance

Source

Cnam-LMSSC

Collection Method

Sensor-based recording using body-conduction transducers with expert and crowdsourced annotation

Freshness

Last updated November 2025; based on research published in 2024.

Geography

France