引導式測驗 · 中等 等級

對話、音樂與音效分離測驗

測試三個電影主幹、混合限制、洩漏和失真、合成到真實的傳輸、歌唱模糊性和來源權。

問題 1 的 8

Which three target stems define the cinematic separation task described here?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. Which three target stems define the cinematic separation task described here?
    • Pitch, tempo and loudness parameters
    • Dialogue, music and sound effects
    • Left, right and center channels
    • Transcript, captions and translations
  2. Why can an estimated dialogue track lose consonants?
    • Consonants never occur in films
    • A separated stem is always the original master
    • The separator may assign overlapping speech energy to another stem
    • Music and speech cannot overlap in frequency
  3. What does mixture consistency alone fail to prove?
    • That estimated tracks can sum to a mix
    • That the input is an audio waveform
    • That all stems share a duration
    • That each sound was assigned to the correct source category
  4. Why should synthetic-mixture results be checked on real films?
    • Mastering, reverberation and layered effects may differ
    • Synthetic data cannot have known reference stems
    • Real films contain no dialogue
    • Mastering does not affect source separation
  5. A model puts singing into the dialogue stem. What issue does this expose?
    • Camera calibration failure
    • A boundary between speech and music categories
    • A grant of soundtrack copyright
    • A guaranteed captioning improvement
  6. Which check is essential before quoting a faint word from separated audio?
    • Only the isolated stem because it sounds cleaner
    • Only a separator score without playback
    • Only text generated from the separated audio
    • Listen to the original mix as well as the estimate
  7. What does an isolated stem not provide by itself?
    • Permission to publish or reuse the underlying sound
    • A candidate track for mixing
    • A way to audition an estimated voice stem
    • A file that contains audio samples
  8. How should accessibility-focused dialogue enhancement be judged?
    • Only the separator’s training loss
    • Only output file size
    • Listener intelligibility in realistic playback, alongside signal measures
    • Only a visualization of music energy