文章摘要
武正平,马建芬,张朝霞,杨东东.改进的正交匹配追踪的语音增强算法*[J].,2018,37(6):934-939
改进的正交匹配追踪的语音增强算法*
Speech enhancement algorithm based on improved orthogonal matching pursuit
投稿时间:2017-12-20  修订日期:2018-11-01
中文摘要:
      为了提高传统正交匹配追踪(Orthogonal Matching Pursuit,OMP )算法的语音增强性能和运算速度,本研究基于稀疏编码理论,提出了一种改进的OMP算法的语音增强算法。其一,将K-奇异值分解(K-singular value decomposition,K-SVD)算法与OMP算法相结合,通过设置能量阈值的方法,提高OMP算法的语音增强性能;其二,通过改进传统OMP算法中信号稀疏逼近的计算方法,提高算法的运算速度。改进的OMP算法的语音增强算法与传统K-SVD语音增强算法相比,采用PESQ评价增强语音的质量,NCM评价语音的可懂度。在NCM的值基本保持不变的情况下,PESQ的值平均提高约12.47%,取得了更好的增强效果。取得了更好的增强效果。改进的OMP算法的运算速度与传统OMP算法相比提高近一倍。
英文摘要:
      In order to improve the running speed of traditional Orthogonal matching pursuit (OMP) algorithm and improve the performance of the enhanced speech. Based on sparse coding theory, this paper proposes an improved speech enhancement algorithm for Orthogonal Matching Pursuit (OMP) algorithm. Firstly, combining K-singular value decomposition (K-SVD) algorithm with OMP algorithm, we set up the energy threshold method to improve the speech enhancement performance of OMP algorithm. Secondly, we improve the algorithm speed by improving the traditional calculation method of signal sparse approximation in OMP algorithm. The speech enhancement algorithm of the improved OMP algorithm is compared with the traditional K-SVD algorithm, the algorithm uses the PESQ to evaluate the quality of the speech, and the NCM is used to evaluate the speech quality. The value of PESQ increased by about 12.47% on average while the value of NCM remained essentially unchanged.This algorithm has achieved better results. The computational speed of the improved OMP algorithm is nearly twice as high as that of the traditional OMP algorithm.
DOI:10.11684/j.issn.1000-310X.2018.06.015
中文关键词: 过完备字典,正交匹配追踪,K-奇异值分解,语音增强
英文关键词: Overcomplete dictionary, Orthogonal matching pursuit, K-singular value decomposition, Speech enhancement
基金项目:山西省重点研发计划(国际合作)项目;山西省自然科学基金
作者单位E-mail
武正平 太原理工大学 1334676753@qq.com 
马建芬 太原理工大学 majianfentyut@126.com 
张朝霞 太原理工大学 zhangzhaoxia1@126.com 
杨东东 太原理工大学 2675411128@qq.com 
摘要点击次数: 88
全文下载次数: 81
查看全文   查看/发表评论  下载PDF阅读器
关闭