Search

A text prompted speaker verification system is presented in this paper. This system is based on ten Chinese digits. Basic acoustic models are speaker dependent and content dependent phoneme HMMs which were generated by adapting speaker independent models to the utterances of specific speakers. An obvious constraint for normalization techniques used in TDSV is that the phrases with the same content should be used for competitive cohort models. So many of the score normalization techniques are either difficult to implement because of lack of data or not good for performance improvement because of poor estimation of the normalization parameters. We propose a method which combines the traditional T-Norm and Cohort Norm together to find a good tradeoff of testing utterance normalization and target speaker model normalization. The proposed method improved the system performance from the baseline equal error rate 3.42% for T-Norm and 2.72% for Cohort Norm to 2.50%. [ABSTRACT FROM AUTHOR]

Utterance verification is a process, in which a spoken utterance is verified against the given keyword. This process is used to make a decision on acceptance or rejection. In this paper, we propose a new approach to the utterance verification, using a boosting classifier with ten confidence measures. This classifier combines a set of 'weak' learners into a 'strong' one. The experimental results present that it can remarkably improve the verification performance. Compared with a single confidence measure, the equal error rate is reduced by up to 23%. The results also show that the boosting classifier is better than the SVM and MLP classifiers, in term of the equal error rate. [ABSTRACT FROM AUTHOR]