| 一种广义边距区分性训练准则 |
| |
| 引用本文： | 陈斌, 张连海, 屈丹, 李弼程. 一种广义边距区分性训练准则[J]. 声学学报, 2014, 39(1): 119-129. DOI: 10.15949/j.cnki.0371-0025.2014.01.011 |
| |
| 作者姓名： | 陈斌 张连海 屈丹 李弼程 |
| |
| 作者单位： | 1.解放军信息工程大学信息系统工程学院 郑州 450002 |
| |
| 基金项目： | 国家自然科学基金(61175017)资助 |
| |
| 摘 要： | 通过分析不同区分性训练目标函数之间的关系,以MMI(Maximum Mutual Information)作为分离度量,把不同的区分性训练目标函数统一为基于广义边距的区分性训练准则.并在该广义边距准则下,通过对其权重函数进行讨论,得到两种区分性训练目标函数:利用组合增进因子和候选词路径中误识词个数,加权候选路径,得到SBMMI(Soft Boosted MMI)目标函数;利用基于单个候选词的后验概率定义每一训练语句的错误识别率,采用幂指数的形式对单个候选词动态加权,得到VWMMI(Variable Weighting MMI)目标函数。实验结果表明,与软边距估计准则和增进的最大互信息方法相比,SBMMI方法准确率分别提高了0.89%和0.56%,VWMMI方法能在SBMMI方法基础上提高0.68%.
|
| 关 键 词： | 目标函数 区分性 训练过程 后验概率 边距 估计准则 识别错误 声学模型 识别性能 准确率 |
| 收稿时间： | 2012-05-14 |
| 修稿时间： | 2013-05-07 |
A discriminative training criteria based on generalized margin |
| |
| CHEN Bin, ZHANG Lianhai, QU Dan, LI Bicheng. A discriminative training criteria based on generalized margin[J]. ACTA ACUSTICA, 2014, 39(1): 119-129. DOI: 10.15949/j.cnki.0371-0025.2014.01.011 |
| |
| Authors: | CHEN Bin ZHANG Lianhai QU Dan LI Bicheng |
| |
| Affiliation: | 1.Institute of Information System Engineering, PLA Information Engineering University Zhengzhou 450002 |
| |
| Abstract: | By analyzing the relationship between different discriminative training objective function and MMI (Maximum Mutual Information) being as the separation measure, the different discriminative training objective function is unified into a discriminative training criteria based on generalized margin. The weighting function in the criteria is further discussed and two kinds of discriminative objective function are got. When the candidate path is weighted through a combination of boosted factor and the number of the misrecognition words in the candidate path, a discriminative objective function SBMMI (Soft Boosted MMI) is presented. While a single candidate word is dynamic weighted using the exponential form in which the misrecognition rate of each training statement is defined by the posterior probability of a single candidate, another discriminative objective function VWMMI (Variable Weighting MMI) is proposed. The experimental results show that compared with the soft margin estimation and boosted maximum mutual information method, the recognition accuracy of SBMMI method increases by 0.89% and 0.56% separately and VWMMI method has a 0.68% improvement upon SBMMI method. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《声学学报》浏览原始摘要信息 |
|
点击此处可从《声学学报》下载全文 |
|
