自适应噪声抵消语声增强的实时后处理算法  

Real-time post-treatment algorithm for adaptive noise cancellation used in speech enhancement

作　　者：连兵[1] 黄杰[1] 王宏禹[1] 

机构地区：[1]大连理工大学信息技术研究所

出　　处：《大连理工大学学报》1996年第1期94-99,共6页Journal of Dalian University of Technology

摘　　要：对自适应噪声抵消系统中两路噪声相关性较弱的情况，采用相于函数作为频域增益因子，对自适应噪声抵消作进一步处理（或后处理），以改善语声增强效果．提出一种实时迭代算法取代现有的基于短时傅氏变换的算法，新算法可简化计算且便于与自适应系统一起实时实现．在计算机模拟实验中，后处理算法的有效性得到了验证．Adaptive noise cancellation systems are widely used in speech enhancement. Under the condition of comparatively weak correlation between the two noises involved, the following two questions are studied: (1) Coherence function is used as a frequency domain amplification factor for the further processing(or post-treatment)of the adaptive noise cancellation system, so as to improve its speech enhancement effect;(2) A real-time recursive algorithm is put forward in substitute for current algorithms based on Short Time Fourier Transform. The new algorithm will simplify computations and will be suited for real-time implementation together with the adaptive system. The effectiveness of the post-treatment algorithm is verified in computer simulations.

关 键 词：自适应滤波 噪声消除 相干函数 语声增强 

分 类 号：TN912.3[电子电信—通信与信息系统]