microsoft/DNS-Challenge resource

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1465 stars · 456 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Microsoft's repository for the Deep Noise Suppression (DNS) Challenge, containing datasets, scripts, and baseline models for training speech enhancement models. It supports noise suppression, de-reverberation, and interfering talker suppression for headset and speakerphone scenarios.

Use cases

• train a deep learning model to suppress background noise in speech
• build a speech enhancement model for headset calls
• denoise meeting recordings with interfering talkers
• evaluate noise suppression models with ITU-T P.835 metrics
• generate synthetic noisy speech training data
• compute speaker embeddings for personalized speech enhancement

When to choose

• you are participating in the DNS Challenge or building speech enhancement models
• you need large-scale noisy/clean speech datasets for training denoising models
• you want standardized P.835-based evaluation of noise suppression quality

When to avoid

• you need a ready-to-use noise suppression product rather than training resources
• you work with non-speech audio like music denoising
• you need real-time noise suppression in a production app without training your own model

Facets

dataset · maturity active

audio-processing machine-learning speech-recognition benchmarking data-generation speech-processing machine-learning artificial-intelligence python cross-platform noise-suppression speech-enhancement dns-challenge dereverberation audio-datasets icassp deep-learning audio

Member repositories

For agents

markdown · JSON · MCP: product_card(name="microsoft/DNS-Challenge")