LibriVAD: A Scalable Open Dataset with Deep Learning Benchmarks for Robust Voice Activity Detection

arXiv:2512.17281 · cs.SD, cs.LG · Submitted 2025-12-19 · Read on arXiv

cs.SD, cs.LG

Submitted: 2025-12-19

Updated: 2026-10-27

Comments: 22 pages. Published in Computer Speech & Language

Journal ref: Computer Speech & Language, vol. 102, Art. 102043, 2027

DOI: 10.1016/j.csl.2026.102043

Code: https://github.com/zhenghuatan/librivad

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

The gist: Robust Voice Activity Detection (VAD) remains a challenging task, especially under noisy, diverse, and unseen acoustic conditions.

Terminology

Abstract

Robust Voice Activity Detection (VAD) remains a challenging task, especially under noisy, diverse, and unseen acoustic conditions. Beyond algorithmic development, a key limitation in advancing VAD research is the lack of large-scale, systematically controlled, and publicly available datasets. To address this, we introduce LibriVAD -- a scalable open-source dataset derived from LibriSpeech and augmented with diverse real-world and synthetic noise sources. LibriVAD enables systematic control over speech-to-noise ratio, silence-to-speech ratio (SSR), and noise diversity, and is released in three sizes (15 GB, 150 GB, and 1.5 TB) with two variants (LibriVAD-NonConcat and LibriVAD-Concat) to support different experimental setups. We benchmark multiple feature-model combinations, including waveform, Mel-Frequency Cepstral Coefficients (MFCC), and Gammatone filter bank cepstral coefficients, and introduce the Vision Transformer (ViT) architecture for VAD. Our experiments show that ViT with MFCC features consistently outperforms established VAD models such as boosted deep neural network and convolutional long short-term memory deep neural network across seen, unseen, and out-of-distribution (OOD) conditions, including evaluation on the real-world VOiCES dataset. We further analyze the impact of dataset size and SSR on model generalization. Experimental results show that scaling the dataset size from small-scale to medium- or large-scale noticeably and consistently improves VAD performance under OOD conditions, while appropriately balancing SSR enhances OOD performance in the small-scale setting. All datasets, trained models, and code are publicly available to foster reproducibility and accelerate progress in VAD research.

Sources

Related papers