UCONV-Conformer: High Reduction of Input Sequence Length for End-to-End Speech Recognition.
Andrei AndrusenkoRauf NasretdinovAleksei RomanenkoPublished in: ICASSP (2023)
Keyphrases
- end to end
- speech recognition
- hidden markov models
- speech processing
- language model
- speech recognizer
- speech signal
- pattern recognition
- noisy environments
- speech synthesis
- automatic speech recognition
- speech recognition systems
- speech recognition technology
- speech retrieval
- speaker identification
- congestion control
- speaker dependent
- speaker independent
- signal processing
- machine learning