计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (26): 144-146.DOI: 10.3778/j.issn.1002-8331.2008.26.044

• 数据库、信号与信息处理 • 上一篇    下一篇

抗噪音谱估计误差干扰的语音增强处理

吴俊伟1,2,何良华1,2   

  1. 1.同济大学 电子与信息工程学院,上海 201804
    2.同济大学 国家高性能计算中心 同济分中心,上海 201804
  • 收稿日期:2008-04-08 修回日期:2008-07-14 出版日期:2008-09-11 发布日期:2008-09-11
  • 通讯作者: 吴俊伟

Speech enhancement technique on poor noise estimation

WU Jun-wei1,2,HE Liang-hua1,2   

  1. 1.Department of Computer Science and Technology,Tongji University,Shanghai 201804,China
    2.Tongji Branch,National Engineering and Technology Center of High Performance Compute,Tongji University,Shanghai 201804,China
  • Received:2008-04-08 Revised:2008-07-14 Online:2008-09-11 Published:2008-09-11
  • Contact: WU Jun-wei

摘要: 在实际应用中通常无法精确估计得到背景噪音谱,传统语音增强效果也随之大大降低。为弱化估计误差引入的干扰,在对数最小均方差估计器(LSA)语音增强方法基础上提出了一套切实可行的增强方案。引入信号检测自动机判别帧成分,针对帧与帧之间的不同特点采取不同级别的噪音抑制处理方案,对确定为噪音帧的部分进行进一步深度抑制,而语音帧部分则沿用改进的LSA方法。实验表明,使用方法能有效抑制背景噪音,特别当噪音谱估计误差较大情况下,相比于LSA该方法具有更优秀的去噪、抗干扰性能。

Abstract: The traditional denoise techniques ability would be quickly descend when it can’t get the precise estimation of the noise spectrum.In this paper,a speech enhance technique is proposed which analyzes the speech components and noise components first and then deals the different frames with different levels of suppression methods,has a supper enhance effect when working under a poor estimation of noise spectrum condition.Extensive objective experiments shows that,after enhancement,the residual noise is further suppressed by the proposed approach comparing with the LSA estimator.