摘要： 从说话人自适应技术讨论了最大后验概率(MAP)和最大似然线性回归(MLLR)两种经典的说话人自适应方法,通过在渐进的MAP方法中引入一个简化的MLLR模块,提出一种适合于强健语音识别的快速综合渐进自适应语音识别方法和策略。
关键词:
语音识别,
自适应,
MAP,
MLLR
Abstract: The paper discusses two classical speaker adaptation methods in the field of speaker adaptation techniques: maximum a posteriori (MAP) and maximum likelihood linear regression (MLLR). By introducing a simplified MLLR module to the incremental MAP processing, a new approach and strategy of speed recognition for rapid, integration incremental adaptation are presented, and the new approach is suited for the sturdy speech ecognition.
中图分类号:
丰洪才,卢正鼎. 基于MAP和MLLR的综合渐进自适应方法研究[J]. , 2005, 31(05): 12-15.
FENG Hongcai1,2, LU Zhengding 1. Integration Incremental Adaption Method Research Based on MAP & MLLR[J]. , 2005, 31(05): 12-15.