引導式測驗 · 中等 等級

音頻測驗中的起始偵測

了解音訊開始偵測器如何發現事件攻擊、使用頻譜通量和峰值拾取以及處理漸進或重疊的聲音。

相關引導路徑音訊中的起始檢測
問題 1 的 8

音訊中的「開始」到底是什麼?

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. 音訊中的「開始」到底是什麼?
    • 聲學事件開始的那一刻,例如音符的起奏
    • 聲音的淡出
    • 曲目的平均響度
    • 一首歌的結尾
  2. 光譜通量測量什麼來偵測起始點?
    • A rule that marks every beat as a note onset
    • Positive increases in spectral energy between successive frames
    • A decay detector that selects only falling energy
    • A pitch estimator that finds the nearest MIDI note
  3. 為什麼在光譜通量中應用半波整流?
    • 放大衰變
    • 將音訊轉換為文字
    • 因此,只有上升(正)能量變化才有貢獻,忽略衰變
    • 減慢音訊速度
  4. 哪種類型的發作最難檢測?
    • 尖銳的鼓聲敲響
    • 一聲響亮的鈸聲
    • 小鼓後拍
    • 柔和、漸進的小提琴聲
  5. 具有自適應閾值的峰值拾取階段可以防止什麼?
    • 雙觸發和假啟動
    • 音訊壓縮
    • 變調
    • 立體聲擴寬
  6. What can a neural onset detector learn from labeled audio examples?
    • A guarantee that every onset will be detected
    • Patterns in the audio features that signal event beginnings
    • A fixed tempo that applies to every recording
    • The identity of every instrument without evaluation
  7. What does onset backtracking do when detected times are used to create audio slices?
    • Moves each onset to the end of the entire recording
    • Replaces each onset with the nearest musical beat
    • Moves an onset to the nearest preceding local minimum of an energy signal
    • Adds a fixed one-second delay after every peak
  8. What can happen when an onset model is transferred to a different instrument or dataset?
    • Its predictions are guaranteed to remain identical
    • The audio automatically becomes monophonic
    • The model no longer needs a time-frequency input
    • Its performance may change and should be evaluated in the new domain