基于状态驻留时间下由转移弧产生输出的HMM模型中的参数估计

本讨论了用状态驻留时间来模型化传统的HMM模型。HMM的一个基本假设是它认为语音信号是准平稳的。然而由状态输出yt的HMM模型，并不能很好地表征语音信号中平稳段或平稳段之间的具体特征；由转移弧产生输出的自左向右HMM系统,则对语音特征作更为细致的描述。本主要讨论在[2]的基础上,对新建模型进行参数估计。

Parameter Estimation Based on State Duration in the Transition-arc Output Hidden Markov Models (HMM)

In this paper , modeling traditional HMM with time-dependent state is discussed. We basically suppose that speech in HMM signal is normally smooth. But state output HMM can not perfectly characterize the smooth segment or the transition interval of two segments. The state transition-arc output HMM can do with the speech more accurately. On the basis of, parameter estimation of new model are mainly discussed .