How Neural Network Depth Compensates for HMM Conditional Independence Assumptions in DNN-HMM Acoustic Models.

Suman V. Ravuri Steven Wegmann

Published in: INTERSPEECH (2016)

Keyphrases

hidden markov models
speech recognition
neural network
independence assumption
acoustic models
automatic speech recognition
discriminative training
pattern recognition
conditional probabilities
speech signal
machine learning
speaker independent
information extraction
maximum likelihood
speech recognizer