摘要
隐马氏模型在语音识别和生物信息学中有重要的应用.本文研究二阶隐马氏模型(HMM2)的基本算法,利用归一化和递推原理,改进模型的前向-后向算法及Baum-Welch训练算法并给予证明,使得该算法更容易理解和机器实现,并保证数值稳定性.将HMM2应用到miRNA靶基因预测的后期过滤处理中取得了较好的结果.
The hidden Markov model has important applications in speech recognition and bioinformatics. This paper studies basic algorithms of the second-order hidden Markov model(HMM2),improves the forward-backward algorithm and Baum-Welch training algorithm of the model.We provide the proof using normalization and recursion,making them easier to be understood and implemented in programming,and ensuring numerical stability.The HMM2 is applied to miRNA target predictions of post-processing filters with good results.
出处
《应用科学学报》
EI
CAS
CSCD
北大核心
2010年第3期307-312,共6页
Journal of Applied Sciences
基金
国家自然科学基金(No.30871341)
上海市重点学科基金(No.S30104)
上海市教委重点学科建设项目基金(No.J50101)
科技部重大科技专项基金(No.2008ZX10002-017,No.2008ZX10002-020,No.2009ZX09103-686)资助