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

TL;DR

This paper introduces LibriVAD, a large-scale, open-source dataset for voice activity detection, and benchmarks deep learning models including a Vision Transformer, demonstrating improved performance and generalization under diverse noisy conditions.

Contribution

The paper presents LibriVAD, a scalable dataset with controlled noise conditions, and evaluates novel deep learning architectures like ViT, advancing VAD research and benchmarking standards.

Findings

ViT with MFCC features outperforms traditional models.

Scaling dataset size improves OOD generalization.

Balancing SSR enhances model robustness.

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, experimentally showing that scaling up dataset size and balancing SSR noticeably and consistently enhance VAD performance under OOD conditions. All datasets, trained models, and code are publicly released to foster reproducibility and accelerate progress in VAD research.