Listen, Adapt, Better WER: Source-free Single-utterance Test-time Adaptation for Automatic Speech Recognition.
Guan-Ting LinShang-Wen LiHung-yi LeePublished in: CoRR (2022)
Keyphrases
- automatic speech recognition
- speech recognition
- word error rate
- speaker adaptation
- speech signal
- conversational speech
- speech segments
- hidden markov models
- language model
- broadcast news
- spoken words
- noisy environments
- adaptation process
- recognition errors
- speech corpus
- speech retrieval
- speech recognizer
- speech synthesis
- pattern recognition
- handwriting recognition
- word recognition
- image processing
- speaker independent
- error rate
- multiscale