UBM Based Speaker Selection and Model Re-Estimation for Speaker Adaptation

摘要：

Based on speaker selection, speaker adaptation technology can get a promising performance. In such system, how to represent a speaker and the computation of selection are still big issues. In this paper, we take Gaussian mixture model (GMM) as representation of a speaker, which adapted from universal background model (UBM). Likelihood ratio (LR) and cross likelihood ratio (CLR) are utilized for speaker selection. Furthermore, a single-pass re-estimation procedure, conditioned on the speaker-independent model is shown.This adaptation strategy was evaluated in a large vocabulary speech recognition task. A relative gain of 11％ with respect to the baseline system is achieved.

关键词： speaker adaptation speaker selection UBM.

作者: Jian Wang Jun Guo Gang Liu Jianjun Lei

作者单位: School of Information Engineering Beijing University of Posts and Telecommunications

会议类型: 国际会议

会议名称: Firth IEEE International Conference on Cognitive Informatics(第五届认知信息国际会议)

会议地点: 北京

会议语种:英文

页码: 856-860

在线出版日期: 2006-07-17（万方平台首次上网日期，不代表论文的发表时间）