首页|期刊简介|投稿指南|分类索引|刊文选读|订阅指南|证明资料|样刊邮寄查询|常见问题解答|联系我们
张开生,赵小芬,王泽,等. 基于总体平均经验模态分解和一步式字典学习联合去噪的语音端点检测算法[J]. 科学技术与工程, 2020, 20(35): 14536-14542.
张开生,zhaoxiaofen,wangze,et al.Voice endpoint detection algorithm based on EEMD and OS-DL joint denoising[J].Science Technology and Engineering,2020,20(35):14536-14542.
基于总体平均经验模态分解和一步式字典学习联合去噪的语音端点检测算法
Voice endpoint detection algorithm based on EEMD and OS-DL joint denoising
投稿时间:2019-12-21  修订日期:2020-08-30
DOI:
中文关键词:  EEMD算法  OS-DL算法  稀疏表示  子带频带方差  端点检测
英文关键词:EEMD algorithm  OS-DL algorithm  sparse representation  subband frequency band variance  endpoint detection
基金项目:陕西省科技计划项目(2017GY-063);陕西省科技计划项目(2017ZDXM-SF-035);陕西省教育厅专项科研计划项目(16JK1100)
           
作者单位
张开生 陕西科技大学
赵小芬 陕西科技大学
王泽 陕西科技大学
宋帆 陕西科技大学
摘要点击次数: 174
全文下载次数: 60
中文摘要:
      针对复杂环境下语音端点检测准确率低下且检测耗时过长的问题,研究一种基于EEMD和OS-DL联合去噪的语音端点检测算法。首先利用EEMD(总体平均经验模态分解)算法对输入语音进行分解得到IMF(本征模式分量),然后使用OS-DL(一步式字典)算法分别对纯净语音信号与噪声信号进行训练,得到纯净语音信号和噪声信号的幅度谱字典,进而对幅度谱进行稀疏表示,利用得到的系数矩阵重新构建出语音信号频谱,将重构出的语音信号频谱经过傅里叶逆变换得到降噪后的语音信号,最后对降噪后的语音信号利用均匀子带频带方差法进行端点检测。实验结果表明:该算法在复杂环境信噪比低于-10dB情况下检测准确率仍可达到85%以上,且平均检测时间缩短至传统端点检测算法的1/3。
英文摘要:
      Aming at the problem of low accuracy and long time-consuming detection of voice endpoints in complex environments,research on a Speech Endpoint Detection Algorithm Based on Joint Denoising of EEMD and OS-DL.The input speech is first decomposed by EEMD (overall average empirical mode decomposition) to obtain IMF (eigenmode component),The OS-DL (one-step dictionary) algorithm is then used to train pure speech signals and noise signals separately,and get dictionary of amplitude spectrum of pure speech signal and noise ,and then sparse representation of the amplitude spectrum,the coefficient matrix is used to reconstruct the spectrum of the speech signal.The inverse Fourier transform of the reconstructed speech signal spectrum to obtain a noise-reduced speech signal,finally, the uniform subband frequency band variance method is used to detect the speech signal after noise reduction.Experimental results show that ,the detection accuracy of the algorithm can reach over 85% when the signal-to-noise ratio of the complex environment is below -10dB, and the average detection time is shortened to 1/3 of the traditional endpoint detection algorithm.
查看全文  查看/发表评论  下载PDF阅读器
关闭
你是第35087330位访问者
版权所有:科学技术与工程编辑部
主管:中国科学技术协会    主办:中国技术经济学会
Tel:(010)62118920 E-mail:stae@vip.163.com
京ICP备05035734号-4
技术支持:本系统由北京勤云科技发展有限公司设计

京公网安备 11010802029091号