Performance Improvement in Speech Recognition by Weighting HMM Likelihood

은닉 마코프 모델 확률 보정을 이용한 음성 인식 성능 향상

  • 권태희 (고려대학교 전자.컴퓨터공학과) ;
  • 고한석 (고려대학교 전자.컴퓨터공학과)
  • Published : 2003.02.01

Abstract

In this paper, assuming that the score of speech utterance is the product of HMM log likelihood and HMM weight, we propose a new method that HMM weights are adapted iteratively like the general MCE training. The proposed method adjusts HMM weights for better performance using delta coefficient defined in terms of misclassification measure. Therefore, the parameter estimation and the Viterbi algorithms of conventional 1:.um can be easily applied to the proposed model by constraining the sum of HMM weights to the number of HMMs in an HMM set. Comparing with the general segmental MCE training approach, computing time decreases by reducing the number of parameters to estimate and avoiding gradient calculation through the optimal state sequence. To evaluate the performance of HMM-based speech recognizer by weighting HMM likelihood, we perform Korean isolated digit recognition experiments. The experimental results show better performance than the MCE algorithm with state weighting.

References

  1. Fundamentals of Speech Recognition L.Rabiner;B.H.Juang
  2. Spoken Language Processing X.Huang;A.Acero;H.W.Hon
  3. Pattern Recognition v.33 An improved maximum model distance approach for HMM-based speech recognition systems A.H.He;S.Kwong;K.F.Man;K.S.Tang
  4. IEEE Transactions on Speech and Audio Processing v.1 no.1 Estimating hidden markov model parameters so as to maximize speech recognition accuracy L.R.Bahl;P.F.Brown;P.V. de Souza;R.L.Mercer
  5. IEEE Transactions on Signal Processing v.40 no.12 Discriminative learning for minimum error classification B.H.Juang;S.Katagiri
  6. IEEE ICASSP-93 Minimum error rate training based on N-best string models W.Chou;C.H.Lee;B.H.Juang
  7. IEEE Transactions on Speech and Audio Processing v.5 no.3 Minimum classification error rate methods for speech recognition B.H.Juang;W.Chou;C.H.Lee
  8. Proceedings of the IEEE v.88 no.8 Discriminant-function-based minimum recognition error rate pattern-recognition approach to speech recognition W.Chou
  9. Speech Communication v.30 An improved approach to robust speech recognition using minimum error classification M.T.Lin;A.Spanias;P.Loizou
  10. Speech Communication v.19 Performance of HMM-based speech recognizers with discriminative state-weights O.W.Kwon;C.K.Un
  11. IEEE Transactions on Speech and Audio Processing v.10 no.4 Improved generalization of MCE parameter estimation with application to speech recognition D.W.Purnell;E.C.Botha