Up nextNext guide
DCASE Challenge for Acoustic Scenes and Events
Audio AI
Audio AI GUIDE
Deep Noise Suppression Challenge
The Deep Noise Suppression (DNS) Challenge, organized by Microsoft, gives researchers shared data and evaluation tasks for speech enhancement.
Listen to this guideAI-narrated · ~2 min
On this page2 min read
Overview
Its editions test perceptual quality and intelligibility under noisy conditions; challenge results do not establish which commercial apps use a method.
Deep Dive
Microsoft began the DNS Challenge in 2020 and held editions at INTERSPEECH and ICASSP. The challenge released speech and noise data, common test sets, and baselines so researchers could compare suppression systems under standardized conditions. The 2022 ICASSP edition included mobile-device scenarios and a personalized track. It combined subjective ITU-T P.835 ratings of speech quality (SIG), background quality (BAK), and overall quality (OVRL) with word accuracy (WAcc), an objective intelligibility measure based on automatic speech recognition. DNS editions therefore pair listening judgments with other measurements rather than replacing all earlier metrics with one score.
Real-World Implementation
Benchmarking speech-enhancement systems on shared noisy-speech test sets
Testing personalized suppression that retains an enrolled target speaker while attenuating other talkers
Evaluating speech quality and intelligibility for conferencing or assistive-audio research
Using released training data and baselines to compare enhancement methods
The Future of Deep Noise Suppression Challenge
Shared benchmarks make results easier to compare, but a benchmark score does not guarantee performance in every room, device, language, or meeting application. Future evaluations can test more realistic microphones, movement, and latency budgets while reporting both perceptual quality and intelligibility.
Putting it into practice
Why it matters
- Access and reach. It improves accessibility through transcription, narration, and voice interfaces.
- Cost and budget. Media teams can ship polished audio faster with smaller budgets.
- Speed and scale. Customer-facing systems can process spoken interactions at larger scale.
Watch out for
- Voice misuse and impersonation risks increase when consent is missing.
- Accuracy can drop across accents, dialects, or noisy environments.
- Synthetic audio can be mistaken for authentic speech without clear labeling.
Next steps
- Obtain explicit consent for voice capture, cloning, and reuse.
- Test quality across diverse speakers and background conditions.
- Define when a human must review or approve outputs.
- Label synthetic audio and keep provenance records for accountability.
Keep Exploring
Free newsletter
Keep up with AI in 3 minutes a day
One short email each weekday with the three AI stories that actually matter. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Deep Noise Suppression Challenge quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Frequently asked questions
What is Deep Noise Suppression Challenge?
The Deep Noise Suppression (DNS) Challenge, organized by Microsoft, gives researchers shared data and evaluation tasks for speech enhancement. Its editions test perceptual quality and intelligibility under noisy conditions; challenge results do not establish which commercial apps use a method.
Who organized the Deep Noise Suppression (DNS) Challenge?
The DNS Challenge was organized by Microsoft and held across editions associated with venues including INTERSPEECH and ICASSP.
What is the main goal of systems built for the DNS Challenge?
The challenge targets real-time speech enhancement, suppressing noise while preserving the speaker's voice.
Why do live speech-enhancement systems constrain algorithmic lookahead?
Lookahead can be useful, but buffering future samples adds delay; real-time systems balance context against their latency budget.
A common technique in DNS entries is to predict a 'mask.' What does the mask do?
A time-frequency mask is multiplied with the noisy spectrum to suppress noise-dominated bins and retain speech.
What data and test conditions did the 2022 DNS Challenge add?
The 2022 edition added mobile-device scenarios to the blind test set and made training and test data fullband at 48 kHz.
Keep learning
Related guides
More guides picked for this topic