A Framework for Practical Multistream ASR

  title={A Framework for Practical Multistream ASR},
  author={Sri Harish Reddy Mallidi and Hynek Hermansky},
Robustness of automatic speech recognition (ASR) to acoustic mismatches can be improved by using multistream architecture. Past multistream approaches involve training large number of neural networks, one for each possible stream combination. During testing phase, each utterance is forward passed through all the neural networks to estimate best stream combination. In this work, we propose a new framework to reduce the complexity of multistream architecture. We show that multiple neural networks… CONTINUE READING