摘要
在分析最小相位群延迟函数特点的基础上,提出了一种改进的汉语连续语音自动切分算法,该算法利用短时能量和类音节平均统计时长构造的最小相位群延迟函数为切分依据,将连续语音切分成一系列类音节单元.实验结果表明,本算法具有96·4%的切分正确率,比原算法提高5·2%.
In this paper,the characteris＇tics of minimum phase group delay functions are analyed, and a new automatic segmentation algorithm is presented. Based on short-term energy and statistical average duration of syllable-like, the minimum phase group delay functions are used for the segmentation. The experimental result by the proposed algorithm comes up to a correct rate of 96.4%,and it exceeds the original segmentation result by 5.2%.
出处
《徐州师范大学学报（自然科学版）》
CAS
2005年第4期45-49,共5页
Journal of Xuzhou Normal University(Natural Science Edition)
基金
江苏省教育厅高校自然科学基金资助项目(99KJB510002)
江苏省"十五"社科基金资助项目(K3-013)
关键词
最小相位函数
统计平均时长
自动语音切分
minimum phase group delay function
statistical average duration
automatic speech segmentation
