Guide-linked quiz · Medium Level
Far-Field Speech Recognition and CHiME-6 Quiz
Test CHiME-6 dinner-party far-field audio, arrays, diarization, overlap, task-year protocols, and close-mic comparison limits.
Question 1 of 8
What makes far-field ASR harder than close-mic recognition?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- What makes far-field ASR harder than close-mic recognition?
- Far microphones cannot record waveforms
- A room always has one speaker
- Distance, reverberation, noise and overlapping voices
- Close microphones change the spoken language
- Which setting supplied the cited CHiME-6 recordings?
- Only studio audiobook readings
- Only synthetic single-speaker clips
- Only outdoor traffic sensors
- Real dinner-party conversations in homes
- What does speaker diarization estimate?
- A speaker’s legal identity from voice alone
- Which speaker turn occurred when
- The exact meaning of every word
- The room’s floor plan
- Why does the presence of close-worn reference microphones not prove far-field product quality?
- The product may have only distant microphones at deployment
- Close microphones always produce worse audio
- Reference channels are never recorded
- Worn mics eliminate the need for testing
- Which error should be reported separately from WER in a multi-speaker room?
- Only whether the word string is fluent
- Only acoustic quality of one channel
- Assigning words to the wrong speaker
- Only a final transcript file size
- What can array beamforming trade in an overlapping conversation?
- It ensures each simultaneous voice is retained
- It removes room reverberation automatically
- It reveals the speakers’ personal identities
- Emphasizing one direction may suppress another speaker
- Why name the CHiME task and year in a score report?
- Every year is identical by definition
- Different editions can use changed data, tracks or rules
- A score is independent of evaluation data
- Only the workshop venue matters
- When distant words remain unclear, what should a transcript do?
- Mark uncertainty instead of inventing fluent text
- Always complete them from topic context
- Assign them to the loudest speaker automatically
- Delete the source recording
Keep testing yourself
More quizzes picked for your level.