Improving Attention-Based End-to-End Speech Recognition by Monotonic Alignment Attention Matrix Reconstruction.
Ziyang ZhuangKun ZouChenfeng MiaoMing FangTao WeiZijian LiWei HuShaojun WangJing XiaoPublished in: ICASSP (2024)
Keyphrases
- end to end
- speech recognition
- hidden markov models
- speech recognition technology
- speech processing
- noisy environments
- speech synthesis
- language model
- speech recognition systems
- speech signal
- audio visual speech recognition
- speaker independent
- congestion control
- machine learning
- noise reduction
- non stationary
- information retrieval