Title :
Improved generalization of MCE parameter estimation with application to speech recognition
Author :
Purnell, Darryl William ; Botha, Elizabeth C.
Author_Institution :
Dept. of Electr., Electron. & Comput. Eng., Pretoria Univ., South Africa
fDate :
5/1/2002 12:00:00 AM
Abstract :
Discriminative training of hidden Markov models (HMMs) using minimum classification error training (MCE) has been shown to work well for certain speech recognition applications. MCE is, however, somewhat prone to overspecialization. This study investigates various techniques which improve performance and generalization of the MCE algorithm. Improvements of up to 10% in relative error rate on the test set are achieved for the TIMIT dataset
Keywords :
error analysis; hidden Markov models; parameter estimation; signal classification; speech recognition; MCE algorithm generalization; MCE parameter estimation; TIMIT dataset; discriminative training; error rate; hidden Markov models; minimum classification error; speech recognition; Acoustic testing; Automatic speech recognition; Error analysis; Error correction; Hidden Markov models; Multi-layer neural network; Neural networks; Parameter estimation; Speech recognition; Viterbi algorithm;
Journal_Title :
Speech and Audio Processing, IEEE Transactions on
DOI :
10.1109/TSA.2002.1011536
