摘要
本文提出了一种基于最小分类错误准则(MCE)的子词权重参数估计算法,通过MCE训练得到子词的权重系数。子词对词级置信度贡献量的研究表明:韵母的确认能力显著好于声母,在置信性能方面比声母更加稳定和可靠,区分能力优于声母。在130个关键词的关键词检测系统实验表明,采用不同子词贡献权重比等贡献权重时等错误率下降3.05%。
A Minimum Classification Error (MCE) criterion based subwords weighting parameters estimation algorithm is proposed in which the sub-word weighting parameters are derived by the MCE training. Investigation of the contribution of different sub-words on the word-level confidence measure show that Finals significantly outperform the Initials with more reliability and stability in confidence performance, and Finals have more discriminative power than those of Initials. Experiment on keyword spotting system with 130 keywords shows that the system with different sub-word weighting contribution achieved a relative Equal Error Rate (EER) reduction of 3.05% compared with the equal weighting contribution case.
出处
《中文信息学报》
CSCD
北大核心
2008年第2期106-109,128,共5页
Journal of Chinese Information Processing
基金
国家863计划资助项目(2006AA010102)
河北省科技厅资助项目(052135147,042135105)
河北省教育厅资助项目(2005340)
关键词
计算机应用
中文信息处理
语音确认
置信度
似然比检验
最小分类错误
computer application
Chinese information processing
utterance verification
confidence measure
likelihood ratio test
MCE