摘要
近年来话者自适应训练(SAT)方法日益受到重视。然而在实际中此方法通常因为部分方差的估计失误而导致识别性能下降。该文提出了一种应用最大后验概率(MAP)估计方差的全新SAT方法,它能够根据后验概率动态地调整模型的方差,从而解决上述问题。在Switchboard数据库上的实验显示,新方法能够显著地提高识别性能,并且有效地提升系统的稳定性。
Recently there has been a growing interest in speaker adaptive training（SAT）. However, errors can often arise when estimating covariance matrices in the original SAT framework due to the lack of observations in some Gauss components. This paper presents a novel approach which applies maximum a posteriori （MAP） covariance-estimating into original SAT. Experimental results in Switchboard corpus demonstrate that the proposed method can deliver significant reductions in word error rate （WER） and raise the robustness of SAT process.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第20期203-204,212,共3页
Computer Engineering
关键词
语音识别
话者自适应
话者自适应训练
MAP
Speech recognition
Speaker adaptation
Speaker adaptive training（SAT）
Maximum a postefiofi（MAP）
