Reducing the Inter-speaker Variance of CNN Acoustic Models Using Unsupervised Adversarial Multi-task Training.
László TóthGábor GosztolyaPublished in: SPECOM (2019)
Keyphrases
- acoustic models
- multi task
- speech recognition
- hidden markov models
- automatic speech recognition
- multi task learning
- speech recognizer
- broadcast news
- discriminative training
- learning tasks
- supervised learning
- speaker independent
- feature selection
- training process
- multi class
- unsupervised learning
- semi supervised
- learning problems
- speech signal
- neural network
- language model
- transfer learning
- pattern recognition
- image classification
- machine learning
- spoken language
- high dimensional
- speaker identification
- speech recognition systems