Page 104 - 201806

第 37 卷 第 6 期 Vol. 37, No.6
2018 年 11 月 Journal of Applied Acoustics November, 2018

⋄ 研究报告 ⋄

改进的正交匹配追踪的语音增强算法 ∗

武正平 1 马建芬 1† 张朝霞 2 杨东东 1

(1 太原理工大学计算机科学与技术学院 晋中 030600)
(2 太原理工大学物理与光电工程学院 晋中 030600)
摘要 为了提高传统正交匹配追踪算法的语音增强性能和运算速度，该研究基于稀疏编码理论，提出了一种
改进的正交匹配追踪算法的语音增强算法。其一，将 K-奇异值分解算法与正交匹配追踪算法相结合，通过
设置能量阈值的方法，提高正交匹配追踪算法的语音增强性能；其二，通过改进传统正交匹配追踪算法中信
号稀疏逼近的计算方法，提高算法的运算速度。结果表明，改进的正交匹配追踪算法的语音增强算法与传统
K-奇异值分解语音增强算法相比，在短时客观可懂度的值基本保持不变的情况下，语音质量客观评估值提高
10.84%，取得了更好的增强效果；改进的正交匹配追踪算法的运算速度与传统正交匹配追踪算法相比提高近
一倍。
关键词 过完备字典，正交匹配追踪，K-奇异值分解，语音增强
中图法分类号: TP391 文献标识码: A 文章编号: 1000-310X(2018)06-0934-06
DOI: 10.11684/j.issn.1000-310X.2018.06.015

Speech enhancement algorithm based on improved orthogonal

matching pursuit

WU Zhengping 1 MA Jianfen 1 ZHANG Zhaoxia 2 YANG Dongdong 1

(1 Taiyuan University of Technology, College of Computer Science and Technology, Jinzhong 030600, China)
(2 Taiyuan University of Technology, College of Physics and Optoelectronlcs, Jinzhong 030600, China)
Abstract In order to enhance the performance of the speech enhancement and the running speed of traditional
orthogonal matching pursuit (OMP) algorithm, based on sparse coding theory, this paper proposes an improved
speech enhancement algorithm for OMP algorithm. Firstly, combining K-singular value decomposition (K-SVD)
algorithm with OMP algorithm, we set up the energy threshold to improve the speech enhancement performance
of OMP algorithm. Secondly, we increase the algorithm speed by improving the traditional calculation method
of signal sparse approximation in OMP algorithm. The results show that compared with the traditional K-
SVD algorithm, the perceptual evaluation of speech quality (PESQ) of the improved OMP algorithm increases
by 10.84% while the value of short-time objective intelligibility (STOI) remains 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.
Key words Overcomplete dictionary, Orthogonal matching pursuit, K-singular value decomposition, Speech
enhancement

2017-12-20 收稿; 2018-04-12 定稿
山西省重点研发计划 (国际合作) 项目 (201603D421013), 山西省自然科学基金项目 (201701D121009)
∗
作者简介: 武正平 (1992- ), 女, 河南信阳人, 硕士研究生, 研究方向: 语音信号处理。
† 通讯作者 E-mail: majianfentyut@126.com