Wav2code: Restore Clean Speech Representations via Codebook Lookup for Noise-Robust ASR.
Yuchen HuChen ChenQiushi ZhuEng Siong ChngPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2024)
Keyphrases
- noisy environments
- automatic speech recognition
- speech recognition
- noisy speech
- speech signal
- speech enhancement
- noise reduction
- background noise
- image noise
- vector quantization
- computationally efficient
- source code
- bag of words
- image representation
- spontaneous speech
- training set
- word error rate
- action recognition
- face recognition
- additive noise
- probabilistic model
- greater robustness
- multiscale