基于单AVS的空间目标语音增强方法  

A speech enhancement method for spatial target using an acoustic vector sensor

作　　者：邹月娴[1] 王鹏[1] 王文敏[1] 

机构地区：[1]北京大学信息工程学院,深圳物联网智能感知技术工程实验室/现代数字信号处理实验室,深圳518055

出　　处：《清华大学学报（自然科学版）》2013年第6期883-887,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金项目(61271309)

摘　　要：空间目标语音增强技术在移动语音应用中具有重要价值。为有效抑制空间干扰语音和背景噪声,采用声学矢量传感器(acoustic vector sensor,AVS),利用其结构的特殊性和各输出分量间的三角函数关系,引入阻塞矩阵,推导目标语音信号的功率谱,从而实现对AVS全向分量的最佳Wiener滤波获得增强语音。仿真实验表明,该方法能够有效抑制空间语音干扰和背景噪声。该方法采用单AVS,计算复杂度低,具备移动设备上的应用优势。Spatial target speech enhancement has important practical value in mobile voice applications.An acoustic vector sensor（AVS） method is given here to suppress spatial interference and background noise.The method combines the unique structure of the AVS and the trigonometric relationships between AVS components in a block matrix filter.The result efficiently estimates the power spectrum of the target speech so an optimal Wiener filter can be implemented.Finally,the enhanced target speech is obtained by passing the omnidirectional signal of the AVS through the optimum Wiener filter.Simulations show that this method effectively suppresses spatial interference and background noise.With only a single AVS and low computational complexity,this method has advantages in mobile voice applications.

关 键 词：声学矢量传感器(AVS) 语音增强 WIENER滤波 阻塞矩阵 

分 类 号：TN912.3[电子电信—通信与信息系统]