Phoneme Recognition Based on Deep Belief Network

To improve the accuracy of phoneme recognition in continuous speech recognition, a new method based on the deep belief network which can extract the posterior probability of phonemeis proposed. Firstly, a deep belief network is pre-trained layer by layer with Restricted Boltzmann Machine, then by adding an output layer called "softmax" to the network, a deep neural network detecting the posterior probability of phoneme can be created. Subsequently, Backward Propagation algorithm is applied to fine-tune the weights discriminatively and make them better at predicting the probability distribution over the states of hidden Markov models. Finally the sequence of the predicted probability distribution is fed into a standard Viterbi decoder. The experiments show that the proposed method has a better performance on the TIMIT dataset than traditional ways.