摘要
该文将多麦克语音增强方法用于单麦克语音增强,给出了一种改进的奇异值分解语音增强方法。该方法首先对输入矩阵进行雅克比奇异值分解,用得到的奇异值矢量构造语音增强滤波器;然后用输入矩阵与滤波器权矢量相乘来构造各路信号;最后采用麦克风阵列波束形成的方法,得到增强后的语音信号。仿真结果表明,该方法能有效地去除加性噪声,并且改善了语音质量。
This paper applies the microphone array speech enhancement method, and proposes a new speech enhancement method to enhancement method the single microphone speech based on Singular Value Decomposition（SVD）. First, for the input matrix, the method adopts Jacobi singular value decomposition to get the singular value vector as the speech enhancement filter. Then the method multiplies the input matrix and the coefficient matrix of the filter to constitute the so called each channel signals. Finally, the method adopts microphone array beamforming method to gain the enhanced speech. The simulation shows that the proposed method can get rid of the addictive noise very well, and improve the speech quality.
出处
《电子与信息学报》
EI
CSCD
北大核心
2008年第2期357-361,共5页
Journal of Electronics & Information Technology
基金
国家自然科学基金(60372082,60172073)
教育部跨世纪优秀人才基金
关键词
语音增强
雅克比奇异值分解
波束形成
Speech enhancement
Jacobi Singular Value Decomposition（SVD）
Beamforming
