Dynamic Acoustic Unit Augmentation With BPE-Dropout for Low-Resource End-to-End Speech Recognition.
Aleksandr LaptevAndrei AndrusenkoIvan PodluzhnyAnton MitrofanovIvan MedennikovYuri MatveevPublished in: CoRR (2021)
Keyphrases
- end to end
- speech recognition
- speech recognition systems
- speech recognizers
- speaker independent
- speech recognizer
- hidden markov models
- language model
- speech signal
- automatic speech recognition
- speaker identification
- speech processing
- pattern recognition
- acoustic models
- speech synthesis
- noisy speech
- noisy environments
- congestion control
- speech recognition technology
- resource management
- machine learning
- speech retrieval
- isolated word
- speaker recognition
- maximum likelihood
- video sequences