计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (17): 222-227.

• 信息与信号处理 • 上一篇    下一篇

基于噪声估计的二值掩蔽语音增强算法

曹龙涛1,李如玮1,鲍长春1,吴水才2   

  1. 1.北京工业大学 电子信息与控制工程学院,北京 100124
    2.北京工业大学 生命学院,北京 100124
  • 出版日期:2015-09-01 发布日期:2015-09-14

Speech enhancement algorithm based on noise estimation of binary masking

CAO Longtao1, LI Ruwei1, BAO Changchun1, WU Shuicai2   

  1. 1.School of Electronic Information and Control Engineering, Beijing University of Technology, Beijing 100124, China
    2.College of Life Science & Bioengineering, Beijing University of Technology, Beijing 100124, China
  • Online:2015-09-01 Published:2015-09-14

摘要: 针对现有的助听器语音增强算法在非平稳噪声环境下,残留大量背景噪声的同时还引入了“音乐噪声”,致使增强语音可懂度和信噪比不理想等问题。提出了一种基于噪声估计的二值掩蔽语音增强算法,该算法利用人耳听觉感知理论,结合人耳的听觉特性和耳蜗的工作机理。采用最小值控制递归平均(Minima-Controlled Recursive Averaging,MCRA)算法获得估计噪声和初步增强语音;将估计噪声和初步增强语音分别通过可以模拟人工耳蜗模型的gammatone滤波器组进行滤波处理,得到各自的时频表示形式;利用人耳的听觉掩蔽特性,计算含噪语音在时频域的二值掩蔽;利用二值掩蔽得到增强语音。实验结果表明:该算法很大程度上去除了谱减法引入的“音乐噪声”,与基于MCRA谱减法相比,增强语音的语言可懂度指数(Speech Intelligibility Index,SII)、主观语音质量评估(Perceptual Evaluation of Speech Quality,PESQ)和信噪比(Signal to Noise Ratio,SNR)都得到了提高。

关键词: 语音增强, 助听器, 噪声估计, 二值掩蔽

Abstract: In order to solve the residual background noise and the musical noise resulted by the existing speech enhancement algorithm for hearing aids, a speech enhancement algorithm based on noise estimation of binary masking is proposed in this paper. The estimated background noise and initial enhanced speech are obtained by using the minima-controlled recursive averaging algorithm. The estimated noise and the initial enhanced speech are processed by gammatone filter and inner cells model and time-frequency representation is obtained. Binary masking of noisy is calculated. The binary masking is used to synthesize enhanced speech by utilizing human auditory masking in time-frequency domain. Experimental results show that the proposed algorithm is compared with the MCRA algorithm, Speech Intelligibility Index(SII), Perceptual Evaluation of Speech Quality(PESQ) and SNR are improved.

Key words: speech enhancement, hearing aids, noise estimate, binary masking