Title :
A Method to Combine HMM and BPNN on Speech Recognition
Author :
Wu-Feng ; Chai-Yi
Author_Institution :
Inst. of Technol., Taipei
Abstract :
Applied the concept of phonic state in Hidden Markov Model to construct the input matrix in BP Neural Network as modeling and recognition, which can decrease their dimension (almost to 1/3 - 1/5) under the same recognition rate. On the one hand, it can save much of memory storage space; on the other hand, it would get more efficiency in calculation. In sum, it has good effects in the application need of real time response situation.
Keywords :
backpropagation; hidden Markov models; neural nets; speech recognition; BP neural network; Viterbi algorithm; hidden Markov model; phonic state; speech recognition; Cybernetics; Electronic mail; Hidden Markov models; Machine learning; Mechanical engineering; Neural networks; Probability distribution; Space technology; Speech recognition; Viterbi algorithm; BPNN; HMM; HMNM; Recognition rate; Viterbi algorithm;
Conference_Titel :
Machine Learning and Cybernetics, 2007 International Conference on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-0973-0
Electronic_ISBN :
978-1-4244-0973-0
DOI :
10.1109/ICMLC.2007.4370458
