计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (18): 137-140.DOI: 10.3778/j.issn.1002-8331.1702-0345

• 模式识别与人工智能 • 上一篇    下一篇

压缩感知与EMD相结合的带噪面罩语音增强

王  霞1,王  丹1,王光艳2,张  艳1   

  1. 1.河北工业大学 电子信息工程学院,天津 300401
    2.天津商业大学 信息工程学院,天津 300134
  • 出版日期:2017-09-15 发布日期:2017-09-29

Noisy face mask speech enhancement combining compressed sensing with EMD

WANG Xia1, WANG Dan1, WANG Guangyan2, ZHANG Yan1   

  1. 1.School of Electronic and Information Engineering, Heibei University of Technology, Tianjin 300401, China
    2.School of Information Engineering, Tianjin University of Commerce, Tianjin 300134, China
  • Online:2017-09-15 Published:2017-09-29

摘要: 针对带噪面罩语音清晰度和可懂度低的问题,提出了一种将压缩感知和经验模式分解(Empirical Mode Decomposition,EMD)相结合的方法来对带噪面罩语音进行增强。首先对带噪面罩语音进行EMD分解得到其本征模式函数信号分量,对其特定本征模式分量进行小波阈值去噪;然后对全部信号分量进行压缩感知,最后重构信号分量得到增强后面罩语音。由实验结果可知,文中提出的方法去噪效果较好,重构误差较小,稳定性较高,有效地实现了面罩语音的增强。

关键词: 带噪面罩语音增强, 压缩感知, 经验模式分解, 小波阈值

Abstract: For the problem of low intelligibility and intelligibility of noisy mask speech, a method combining compressed sensing and Empirical Mode Decomposition (EMD) is proposed to enhance the performance of noisy mask speech. Firstly, the noisy mask speech is decomposed by EMD to obtain the intrinsic mode function signal components, and specific intrinsic mode component are denoised by wavelet threshold. Then all signal components are compressed sensing and reconstructed to enhance the noisy mask speech. The experimental results show that the proposed method has good denoising effect, small reconstruction error and high stability, and effectively realizes the enhancement of mask speech.

Key words: noisy mask speech enhancement, compressed sensing, Empirical Mode Decomposition(EMD), wavelet threshold