引導式測驗 · 中等 等級

多聲部音訊的旋律提取測驗

測試主要音高與完整樂譜或來源主幹、諧波與八度誤差、聲音、輪廓選擇及參考評估

問題 1 的 8

多聲部歌曲的旋律提取通常輸出什麼?

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. 多聲部歌曲的旋律提取通常輸出什麼?
    • 每件樂器的音符以完整樂譜形式
    • 一個隨時間變化的主要音高輪廓
    • 僅隔離的主唱波形
    • 一個不帶時間標記的和弦序列
  2. 為什麼強烈的諧波會造成八度誤差?
    • 諧波不會出現在音樂中
    • 八度誤差只由檔名造成
    • 每個泛音的頻率與基頻相同
    • 模型可能選擇泛音而非較弱的基頻
  3. What does an unvoiced decision represent in melody tracking?
    • No target melody pitch should be asserted for that interval
    • The recording contains no sound of any kind
    • All instruments have stopped permanently
    • The model has found a new key signature
  4. Why might a guitar solo confuse a vocal-melody tracker?
    • Guitar sound has no pitches
    • The source file loses its sample rate
    • The predominant lead source can change within the song
    • Every song has only one instrument
  5. Which error can make a smooth contour still inaccurate?
    • Choosing a stable sample rate
    • Inventing pitches through rests or unvoiced regions
    • Using a reference melody
    • Reporting octave errors separately
  6. What further step is needed before a pitch curve becomes a usable score?
    • Determine note boundaries, rhythm and other symbolic details
    • Rename the file as MIDI
    • Treat every frame as a separate written note
    • Discard the original audio
  7. Which evaluation set tests transfer beyond clean vocal pop?
    • Only near-duplicates of training songs
    • Only an isolated monophonic example
    • Only the easiest steady vocal verse
    • Songs with solos, varied genres and dense accompaniment
  8. Why preserve human review of extracted melody?
    • Automated contours are always exact
    • A pitch line proves all chords
    • The intended lead and octave can be ambiguous from the mix
    • Listeners cannot detect voicing errors