摘要
在经典隐马氏模型中,假设状态转移概率只和当前状态的临近状态有关,而和以前的状态无关;在t时刻输出观测值的概率只和当前状态有关,而和以前的状态无关.这样的假设不是很合理,因为任一时刻的观测值不仅和当前状态有关,还和以前的状态有关.由此提出了二阶隐马氏模型的基于最大互信息的参数估计算法.
In the classical HMM we suppose that the probability of station transfer is only relevant with the adjacent station, and the observation probability at time t is only relevant with the hidden station at time t. The supposition is not reasonable because in fact the observation at every time is not only relevant with the hidden station at time t, but also relevant with the hidden station at time t-1. In this paper, We put forward a new algorithm of parameters estimation to the new HMM based on maximum mutual information
出处
《伊犁师范学院学报（自然科学版）》
2009年第3期9-12,共4页
Journal of Yili Normal University：Natural Science Edition
关键词
隐马氏模型
参数估计
最大互信息
HMM
estimation algorithm
maximum mutual information
