Singular value decomposition based low-footprint speaker adaptation and personalization for deep neural network.
Jian XueJinyu LiDong YuMike SeltzerYifan GongPublished in: ICASSP (2014)
Keyphrases
- singular value decomposition
- neural network
- speaker adaptation
- singular values
- least squares
- low rank
- latent semantic indexing
- dimensionality reduction
- principal component analysis
- speech recognition
- recommender systems
- pattern recognition
- semi supervised
- hidden markov models
- low rank approximation
- computer vision
- co occurrence
- automatic speech recognition
- singular vectors