Improving the Performance of Acoustic-to-Articulatory Inversion by Removing the Training Loss of Noncritical Portions of Articulatory Channels Dynamically.
Qiang FangPublished in: INTERSPEECH (2020)
Keyphrases
- acoustic features
- speech recognition
- vocal tract
- acoustic models
- formant frequencies
- multi stream
- training set
- speech synthesis
- test set
- training phase
- decision trees
- training samples
- speech recognition systems
- underwater acoustic
- speech signal
- speaker verification
- automatic speech recognition
- real time
- training process
- training examples
- image reconstruction
- source localization
- non stationary
- supervised learning
- neural network