Deep Noise Suppression Challenge Quiz

Answers and explanations appear as you take the quiz.

1. Who organized the Deep Noise Suppression (DNS) Challenge?
   • Microsoft
   • Google
   • Apple
   • Meta
2. What is the main goal of systems built for the DNS Challenge?
   • Translate speech between languages
   • Remove background noise from speech in real time
   • Generate synthetic singing voices
   • Compress audio files for storage
3. Why do live speech-enhancement systems constrain algorithmic lookahead?
   • To prevent the model from using any learned context
   • To make the audio mono
   • To keep end-to-end delay within a usable latency budget
   • To eliminate the need for a noise estimate
4. A common technique in DNS entries is to predict a 'mask.' What does the mask do?
   • It blurs the entire spectrum equally
   • It adds reverberation to the signal
   • It converts audio into text
   • It scales each time-frequency bin to attenuate noise and keep speech
5. What data and test conditions did the 2022 DNS Challenge add?
   • Fullband 48-kHz data and mobile-device scenarios
   • Only telephone-band audio at 8 kHz
   • Video captions and image labels
   • A test set with no speech or background noise
6. What is a 'personalized' DNS scenario?
   • Suppressing all sound including the target speaker
   • Keeping only an enrolled target speaker and removing other voices and noise
   • Adding the user's name to the audio metadata
   • Translating the speaker's accent
7. In the 2022 ICASSP DNS Challenge, what did the combined ranking score average?
   • Only PESQ and file size
   • Speaker identity accuracy and sample rate
   • WAcc and subjective P.835 SIG, BAK, and OVRL scores
   • MOS from a single listener only
8. What does WAcc measure in the 2022 DNS Challenge?
   • The loudness of background noise
   • The number of audio channels
   • The model’s parameter count
   • Word accuracy from automatic speech recognition