LRTD: A Low-rank Transformer with Dynamic Depth and Width for Speech Recognition.
Fan YuWei XiZhao YangZiye TongJingtong SunPublished in: IJCNN (2022)
Keyphrases
- speech recognition
- low rank
- hidden markov models
- missing data
- matrix factorization
- speech synthesis
- language model
- low rank matrix
- automatic speech recognition
- matrix completion
- speech signal
- speech recognition technology
- rank minimization
- convex optimization
- semi supervised
- speech recognition systems
- singular value decomposition
- speech recognizer
- pattern recognition
- linear combination
- speaker independent
- speaker identification
- trace norm
- model selection
- support vector