analiza aiqbench
The paper releases an open speech dataset (LibriVAD) in three sizes and benchmarks deep learning models for voice activity detection, reporting that a Vision Transformer with MFCC features generalises better than established baselines on unseen and real-world data. For readers, the main value is the controlled, publicly available training resource and its finding that dataset scale and speech-to-silence balance affect out-of-distribution performance. The gains are measured on specific benchmarks, so their reach to other deployment conditions still needs independent testing.
LibriVAD: A Scalable Open Dataset with Deep Learning Benchmarks for Robust Voice Activity Detection
arXiv:2512.17281v2 Announce Type: replace-cross 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 \textbf{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.
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 \textbf{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.
22 pages. Published in Computer Speech & Language
Sound (cs.SD); Machine Learning (cs.LG)
arXiv:2512.17281 [cs.SD]
(or arXiv:2512.17281v2 [cs.SD] for this version)
https://doi.org/10.48550/arXiv.2512.17281 Focus to learn more arXiv-issued DOI via DataCite
Computer Speech & Language, vol. 102, Art. 102043, 2027
https://doi.org/10.1016/j.csl.2026.102043 Focus to learn more DOI(s) linking to related resources
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 351 cuvinte
Subiecte conexe
Detect, Explain, Interpret: An End-to-End Benchmark for Time Series Anomaly Detection, Explainability and Interpretability
arXiv · 2 oct. 2026
SALUS: Automated Auditing of NL-to-SQL Benchmarks through Weak Supervision of Multi-Agent Output
arXiv · 6 oct. 2026
LongNovel: A Multi-Scale Benchmark for Hallucination Detection in Long-Context Novel Summarization
arXiv · 20 aug. 2026
Benchmarking Deep Learning Models for Anomaly Detection in Real-World Cloud Telemetry
arXiv · 14 oct. 2026