Aurora-4

Aurora-4 is a dataset used to evaluate automatic speech recognition (ASR) systems, containing noisy speech data designed to test the robustness of ASR models in challenging acoustic conditions.

Papers using Aurora-4 (3)

Multiple-hypothesis RNN-T Loss For Unsupervised Fine-tuning And Self-training Of Neural Transducer2022 · 1 citesSpeech-enhanced and Noise-aware Networks for Robust Speech Recognition2022Multiple-hypothesis RNN-T Loss for Unsupervised Fine-tuning and Self-training of Neural Transducer2022