论文部分内容阅读
语音识别的互信息匹配模型是依据互信息理论而提出的。在互信息匹配理论的现有成果的基础上,提出了基于互信息估计的连续语音识别算法—MI_OneStage算法,实验表明所提出的算法简单有效,优于传统的OneStage算法。
Mutual information matching model of speech recognition is based on mutual information theory. Based on the existing results of Mutual Information Matching Theory, we propose an MI_OneStage algorithm based on Mutual Information Estimation. The experiments show that the proposed algorithm is simple and effective and is superior to the traditional OneStage algorithm.