基于因子分析的隐马尔可夫模型

状态输出概率密度为对角协方差矩阵高斯分布的隐马尔可夫模型(HMM-DG)在帧内特征相关建模方面存在缺陷．本文将因子分析方法与HMM-DG的混合高斯建模相结合，提出了一种具有弹性的帧内特征相关隐马尔可夫模型框架一基于因子分析的隐马尔可夫模型(HMM-FA)．并导出了HMM-FA的训练算法．理论分析和仿真实验都表明：在训练数据相同的条件下，HMM-FA的性能优于HMM-DG。

A hidden Markov model based on factor analysis

The hidden Markov model based on diagonal covariance matrix Gaussian distributions (HMM-DG) is at present the most popular and successful model in speech recognition.However,there are well-known shortcomings in HMM-DG particularly in the modeling of the correlation among feature-vector elements.This paper investigates the combined use of mixture Gaussian models and factor analysis in HMM.We propose a hidden Markov model based on factor analysis (HMM-FA) and derive an expectation-maximization(EM) algorithm for maximum likelihood estimation.Our theoretical analysis and computer simulation show that the HMM-FA can achieve better performances over HMM-DG with the same amount of training data.