Training deep neural networks with non-uniform frame-level cost function for automatic speech recognition.
Aldonso BecerraJosé Ismael de la Rosa VargasEfrén GonzálezÁngel David Pedroza RamírezN. Iracemi EscalantePublished in: Multim. Tools Appl. (2018)
Keyphrases
- automatic speech recognition
- cost function
- neural network
- speech recognition
- training process
- discriminative training
- hidden markov models
- speech retrieval
- word error rate
- speech signal
- conversational speech
- broadcast news
- recognition errors
- artificial neural networks
- pattern recognition
- noisy environments
- acoustic models
- spoken words
- word recognition
- back propagation
- multi modal
- machine learning