首页 | 本学科首页   官方微博 | 高级检索  
     

骨导语音库的建立与骨气导语音的互信息分析
引用本文:邢益搏,张雄伟,郑昌艳,曹铁勇.骨导语音库的建立与骨气导语音的互信息分析[J].声学技术,2019,38(3):312-316.
作者姓名:邢益搏  张雄伟  郑昌艳  曹铁勇
作者单位:陆军工程大学指挥控制工程学院
基金项目:国家自然科学基金资助项目(61471394、61402519)
摘    要:首先设计了适用于骨导语音增强的语料采集方案,采集了1 320句涵盖音节全面的语料,并制定了相应的录音规范;其次介绍了骨导语音库建立的意义,说明了语音库建立的实施方案,建成了由40个说话人录制的包括气导语音和骨导语音各8 000句的语音库;然后在对比骨导语音与气导语音声学特性的基础上,分析了骨气导语音在高频和低频的互信息量,为骨导语音的增强提供了理论依据;最后基于现阶段的研究及文中构建的语音库对今后的研究做出展望。

关 键 词:骨导语音|语音库|互信息分析|语音增强
收稿时间:2018/1/8 0:00:00
修稿时间:2018/2/20 0:00:00

Establishment of bone-conducted speech database and mutual information analysis between bone and airconducted speeches
XING Yi-bo,ZHANG Xiong-wei,ZHENG Chang-yan and CAO Tie-yong.Establishment of bone-conducted speech database and mutual information analysis between bone and airconducted speeches[J].Technical Acoustics,2019,38(3):312-316.
Authors:XING Yi-bo  ZHANG Xiong-wei  ZHENG Chang-yan and CAO Tie-yong
Affiliation:The Army Engineering University of PLA, Institute of Command and Control Engineering, Nanjing 210007, Jiangsu, China,The Army Engineering University of PLA, Institute of Command and Control Engineering, Nanjing 210007, Jiangsu, China,The Army Engineering University of PLA, Institute of Command and Control Engineering, Nanjing 210007, Jiangsu, China and The Army Engineering University of PLA, Institute of Command and Control Engineering, Nanjing 210007, Jiangsu, China
Abstract:In this paper, a corpus acquisition scheme suitable for bone-conducted speech enhancement is designed, total 1 320 syllabic balanced sentences of covering comprehensive syllables are collected and a corresponding recording specification is developed. The significance of establishing bone-conducted speech database and the implementation scheme of the database are introduced, and a database containing 8 000 air-conducted and bone-conduced speeches spoken by 40 speakers is constructed. Based on the comparison of acoustic characteristics between air-conducted and bone-conducted speeches, the mutual information contents between bone and air conducted speeches at high and low frequencies are analyzed, which provides a theoretical basis for the enhancement of bone-conducted speech. Finally, based on the current stage of research and combining the database constructed in this paper, the future research direction is prospected.
Keywords:bone-conducted speech|speech database|mutual information analysis|speech enhancement
本文献已被 CNKI 等数据库收录!
点击此处可从《声学技术》浏览原始摘要信息
点击此处可从《声学技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号