Bài kiểm tra có liên kết hướng dẫn · Cứng Cấp độ

Mô hình âm thanh GMM-HMM trong bài kiểm tra nhận dạng giọng nói

Kiểm tra các trạng thái và chuyển tiếp ẩn HMM, phát xạ âm thanh GMM, từ vựng, giải mã Viterbi, giới hạn mô hình và so sánh thần kinh.

Đường dẫn hướng dẫn liên quanMẫu âm thanh Gmm Hmm
Câu hỏi 1 của 8

In a GMM-HMM recognizer, what does an HMM state commonly represent?

Xem trước mọi câu hỏi8 câu hỏi

Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.

  1. In a GMM-HMM recognizer, what does an HMM state commonly represent?
    • An exact audio-file checksum
    • A hidden phonetic or related speech state in a time sequence
    • The speaker’s permanent identity
    • A directly observed written letter
  2. What does the GMM component contribute to a candidate HMM state?
    • A guaranteed word spelling
    • A new acoustic recording
    • A language-independent identity certificate
    • A likelihood for the observed acoustic feature vector
  3. Why are state transitions useful in speech recognition?
    • Speech sounds occur in structured sequences over time
    • Every frame can be decoded independently
    • The lexicon becomes unnecessary
    • Word boundaries are directly observed
  4. How does a pronunciation lexicon help the conventional pipeline?
    • It measures emotion from audio
    • It replaces acoustic evidence with spelling
    • It connects candidate words with sound-state sequences
    • It guarantees every rare name is covered
  5. What does Viterbi-style decoding search for?
    • The certainly correct sentence
    • A high-probability hidden-state path through the sequence
    • A new image region
    • A microphone hardware setting
  6. Which assumption can limit a simple HMM?
    • The next state depends on limited previous-state context
    • All speakers have identical acoustics
    • Mixture components cannot have weights
    • Words do not require pronunciation models
  7. Why test on speakers and rooms absent from training?
    • A lexicon predicts every room condition
    • Training loss guarantees transfer
    • All speakers produce the same features
    • Clean in-source scores may not transfer across acoustics
  8. What distinguishes a GMM emission model from HMM transitions?
    • Both factors only count written words
    • Transitions score microphone brightness
    • Emissions score observed features under states; transitions score state changes
    • Emissions decide a school timetable