Confidence Measure Based Incremental Adaptation for Online Language Identification

This paper proposes an novel two-pass adaptation method for online language identification by using confidence measure based incremental language model adaptation. In this system, we firstly used semi-supervised language model adaptation to solve the problem of channel mismatch, and then used unsupervised incremental adaptation to adjust new language model during online language identification. For robust adaptation, we compare three confidence measures and then present a new fusion method with Bayesian classifier. Tested on the RMTS(Real-world Multi-channel Telephone Speech) database, experiments show that using semi-supervised language model adaptation, the target language detection rate rises from 73.26% to 80.02% and after unsupervised incremental language model adaptation, an extra rise over 3.91% (from 80.02% to 83.93%) is obtained.

[1]  Liu Jia,et al.  Language identification using PPRLM with confidence measures , 2004, Proceedings 7th International Conference on Signal Processing, 2004. Proceedings. ICSP '04. 2004..

[2]  Rong Zhang,et al.  Word level confidence annotation using combinations of features , 2001, INTERSPEECH.

[3]  Bhiksha Raj,et al.  A boosting approach for confidence scoring , 2001, INTERSPEECH.

[4]  Brian Roark,et al.  Unsupervised language model adaptation , 2003, 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)..

[5]  Hanseok Ko,et al.  Bayesian fusion of confidence measures for speech recognition , 2005, IEEE Signal Process. Lett..

[6]  Y.K. Muthusamy,et al.  Reviewing automatic language identification , 1994, IEEE Signal Processing Magazine.

[7]  Yingna Chen,et al.  Language model adaptation and confidence measure for robust language identification , 2005, IEEE International Symposium on Communications and Information Technology, 2005. ISCIT 2005..

[8]  Marc A. Zissman,et al.  Automatic language identification , 2001, Speech Commun..

[9]  Timothy J. Hazen,et al.  Word and phone level acoustic confidence scoring , 2000, 2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100).

[10]  Douglas A. Reynolds,et al.  Language identification using Gaussian mixture model tokenization , 2002, 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing.