Mismatched training data enhancement for automatic recognition of children's speech using DNN-HMM.
Mengjie QianIan McLaughlinWu QuoLi-Rong DaiPublished in: ISCSLP (2016)
Keyphrases
- automatic recognition
- speech recognition
- training data
- hidden markov models
- training process
- acoustic models
- speech recognition technology
- speech signal
- hearing impaired
- multi stream
- automatic speech recognition systems
- automatic speech recognition
- speaker independent
- speech recognizer
- speech synthesis
- recognition engine
- data sets
- autistic children
- test data
- classification accuracy
- keyword spotting
- test set
- speech processing
- school children
- training set
- decision trees
- pilot study
- visual speech
- speaker adaptation
- young children
- prior knowledge
- audio visual
- speech recognition systems
- learning algorithm
- pattern recognition
- phoneme recognition
- supervised learning
- neural network
- labeled data
- language model
- noisy environments
- training examples
- image enhancement
- gesture recognition
- support vector machine
- semi supervised
- training samples
- character segmentation
- finite state transducers
- speaker identification
- children learn
- vocal tract
- computer games
- license plate
- dialogue system
- text to speech