引用本文: | 韩广军,罗强,李兵.基于DCPM-FGA的隐马氏模型的参数训练[J].国防科技大学学报,2004,26(4):112-114.[点击复制] |
HAN Guangjun,LUO Qiang,LI Bing.Parameters Training of HMM Based on DCPM-FGA[J].Journal of National University of Defense Technology,2004,26(4):112-114[点击复制] |
|
|
|
本文已被:浏览 5780次 下载 5108次 |
基于DCPM-FGA的隐马氏模型的参数训练 |
韩广军, 罗强, 李兵 |
(国防科技大学 理学院,湖南 长沙 410073)
|
摘要: |
建立了一种将直接比较比例方法与模糊遗传算法结合在一起用于隐马氏模型参数训练的方法。针对不同的码本大小进行了仿真试验。结果表明,在码本较大时该方法效果要好于B-W算法和Viterbi算法,码本较小时两种方法性能相近。 |
关键词: 隐马氏模型 直接比较比例方法-模糊遗传算法 码本 |
DOI: |
投稿日期:2004-02-16 |
基金项目:国家自然科学基金资助项目(60375023);国防科技大学基础研究资助项目 |
|
Parameters Training of HMM Based on DCPM-FGA |
HAN Guangjun, LUO Qiang, LI Bing |
(College of Science, National Univ. of Defense Technology, Changsha 410073, China)
|
Abstract: |
Using DCPM-FGA Direct comparison-proportional method and Fuzzy genetic algorithm to train the parameters of HMM is a very good idea. Different simulations are made for different codebook sizes. The results of simulations show that when the codebook size is bigger, the DCPM-FGA is better than B-W algorithm, otherwise, it has almost the same performance with the above algorithms. |
Keywords: HMM DCPM-FGA codebook |
|
|