摘要
提出了一个基于最小统计及谱减法的语音增强方法。不采用端点检测,在语音帧内及噪声帧内都进行噪声更新。为了实时应用的需求,提出了一种加快最小值更新的方法,实验表明,该方法能有效削弱变电平噪声。
in this paper, a speech enhancement approach using minimum estimate and spectral subtraction is proposed. The noise estimation is updated in both speech frame and noise frame without the voice activity detection. A minima searching method is proposed for real-time application. The simulating results show that this algorithm works efficiently.
出处
《电声技术》
2006年第12期43-46,共4页
Audio Engineering
关键词
最小统计
最优平滑
谱减法
语音增强
minimum estimate
optimal smoothing
spectral subtraction
speech enhancement