摘要
根据传统的谱相减语音增强算法,提出了一种抑制噪声的迭代多带谱相减语音增强方法,暨将迭代方法和多带谱相减方法相结合。同时该方法采取了一种不需要静音检测(VAD)的自适应方法来估计噪声。仿真结果表明, 该方法在较好的抑制背景噪声和残留音乐噪声的同时,保持较好的语音可懂度。
According to the traditional spectral subtraction speech enhancement algorithm, a speech enhancement method based on iterative multi- band spectral subtraction algorithm has been put forward to suppress noise. It combines iterative spectral subtraction with multi - band spectral subtraction algorithm. At the same time, this method adopts adaptive noise estimation which does results shows its excellent performance in background audible distortion. not need voice activity detector （VAD）. The simulation noise and music noise suppression while keeping very low
出处
《西安邮电学院学报》
2005年第1期15-18,共4页
Journal of Xi'an Institute of Posts and Telecommunications
关键词
语音增强
自适应噪声估计
多带谱相减
speech enhancement
adaptive noise estimation
multi- band spectral subtraction
