Exploiting Cross Domain Acoustic-to-Articulatory Inverted Features for Disordered Speech Recognition.
Shujie HuShansong LiuXurong XieMengzhe GengTianzi WangShoukang HuMingyu CuiXunying LiuHelen MengPublished in: ICASSP (2022)
Keyphrases
- speech recognition
- speech recognition systems
- cross domain
- hidden markov models
- speech recognizers
- speaker independent
- speech synthesis
- speech signal
- pattern recognition
- language model
- speech recognizer
- knowledge transfer
- noisy speech
- mel frequency cepstral coefficients
- automatic speech recognition
- sentiment classification
- feature vectors
- noisy environments
- vocal tract
- transfer learning
- computer vision
- feature space
- e learning
- speaker identification
- training data
- feature set
- prior knowledge
- neural network