摘要： 建立一个基于框架的关键词识别系统,其第一步工作是要对每个关键词建立模型,如何选取适当的训练样本数训练出一套好的HMM模型参数将直接影响到该关键词识别系统的性能好坏。比较了使用不同的样本数训练出来的多套模型的性能,并通过使用不同的模型进行搭配得到了最佳的识别率。
关键词:
隐含马尔可夫模型,
关键词确认,
训练样本
Abstract: The training of keyword models is a pivotal problem for HMM based keyword spotting systems. In this paper we compare the performance of different keyword model sets trained with different number of samples. By studying the results we manage to gain the best performance. ;;
中图分类号:
欧嘉致,刘丙哲,陈洪,李宗葛. 基于HMM框架的关键词确认系统中关键词训练样本数的研究[J]. , 2001, 27(07): 121-122.
OU Jiazhi, LIU Bingzhe, CHEN Hong, LI Zongge. Training Keyword Model for HMM Based Word Spotting System with Multiple Number of Samples[J]. , 2001, 27(07): 121-122.