On the compression of recurrent neural networks with an application to LVCSR acoustic modeling for embedded speech recognition.
Rohit PrabhavalkarOuais AlsharifAntoine BruguierIan McGrawPublished in: ICASSP (2016)
Keyphrases
- speech recognition
- recurrent neural networks
- speech recognition systems
- speaker independent
- echo state networks
- speech recognizers
- hidden markov models
- neural network
- acoustic models
- speech synthesis
- speech processing
- speech recognizer
- feed forward
- language model
- automatic speech recognition
- speech signal
- pattern recognition
- noisy environments
- speech recognition technology
- artificial neural networks
- recurrent networks
- speech retrieval
- noisy speech
- speaker dependent
- speaker recognition
- speaker identification
- image compression
- speaker verification
- acoustic features
- signal processing
- feature selection