DNSMOS

DNSMOS is a dataset that contains perceptual mean opinion scores (MOS) for evaluating speech quality, used to provide pseudo-labels in the context of speech quality assessment.

Papers using DNSMOS (6)

Towards Robust Generative Speech Enhancement Using Vector Quantisation-Based Neural Audio Codec2026Few-Shot and Pseudo-Label Guided Speech Quality Evaluation with Large Language Models2026When Fine-Tuning Fails and when it Generalises: Role of Data Diversity and Mixed Training in LLM-based TTS2026Reducing the Gap Between Pretrained Speech Enhancement and Recognition Models Using a Real Speech-Trained Bridging Module2025A Study on Zero-shot Non-intrusive Speech Assessment using Large Language Models2024Distil-DCCRN: A Small-footprint DCCRN Leveraging Feature-based Knowledge Distillation in Speech Enhancement2024 · 2 cites